{"about":{"non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","site":"https://codewithpapers.app","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page","syntology":{"site":"https://syntology.ai","developers":"https://syntology.ai/developers","mcp":{"server":"https://syntology.ai/mcp","transport":"streamable-http","server_card":"https://syntology.ai/.well-known/mcp/server-card.json","auth":{"type":"trial token, no account","trial_token":"https://syntology.ai/api/oauth/trial/token","method":"POST","docs":"https://syntology.ai/developers"}},"have":"https://syntology.ai/api/graph/have?x=<method, arXiv id or title> (free, answers coverage only)","paper_base":"https://syntology.ai/paper/","atlas_base":"https://app.syntology.ai/?focus="},"machine_readable":[{"url":"https://codewithpapers.app/llms.txt","what":"the machine catalog: every machine-readable file, counted"},{"url":"https://codewithpapers.app/index/manifest.json","what":"paper-to-code index by arXiv id, with Syntology's counts"},{"url":"https://codewithpapers.app/search/manifest.json","what":"site search index (titles, authors) and its files"},{"url":"https://codewithpapers.app/download","what":"bulk files: Syntology's layer, described there"},{"url":"https://codewithpapers.app/build_manifest.json","what":"the build record: inputs, counts, exclusions, probes"}]},"url":"/task/safety-alignment","name":"Safety Alignment","slug":"safety-alignment","description_markdown":null,"categories":[{"name":"Natural Language Processing","url":"/area/natural-language-processing"}],"source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","slug_source":"archive_url"},"counts":{"papers_tagged":288,"papers_with_code":134,"benchmarks":0,"benchmark_tables_in_archive":0,"benchmark_tables_shown":0,"benchmark_tables_withheld_as_spam":0,"benchmark_definition":"a leaderboard table with at least one row; benchmark_tables_shown also counts the zero-row tables; benchmark_tables_in_archive adds the tables withheld as spam","datasets":1,"subtasks":0,"parent_tasks":0},"benchmarks":[],"datasets":[{"url":"/dataset/sudo-dataset","name":"SUDO Dataset","full_name":"","num_papers_in_archive":1}],"subtasks":[],"parent_tasks":[],"papers":{"order":"repositories listed in the archive (most first), then date (newest first), then slug; the archive holds no stars","population":"every paper the archive tags with this task (distinct papers; spam titles not listed). 'of' counts the papers listed; until 2026-09-25 it counted only the papers with a code link, and that number is of_with_code","page":1,"pages_in_order":3,"rows_per_page":100,"shown":100,"of":288,"of_with_code":134,"tagged_in_all":288,"counts":{"archive_papers_tagged":288,"with_a_code_link":134,"where_syntology_ran_a_sample":73,"not_listed_spam_title":0,"listed":288,"listed_where_code_ran":73,"where_syntology_ran_a_sample_split":{"with_a_run_with_no_instrument_failure":54,"every_run_a_failure_of_syntologys_instrument":19,"listed_with_a_run_with_no_instrument_failure":54,"listed_every_run_a_failure_of_syntologys_instrument":19,"filter":{"states":["a run with no instrument failure","any run, instrument failures included"],"default":"a run with no instrument failure","note":"on the 'only where code ran' pages the default hides, in the browser, the rows where every run was a failure of Syntology's instrument; the second state shows them again. Rows are hidden, never re-ordered; these twins list every row"}},"definition":"distinct papers the archive tags; 'where Syntology ran a sample' counts papers with at least one harvested sample that ran, which is not a correctness claim"},"items":[{"url":"/paper/the-hidden-dimensions-of-llm-alignment-a","slug":"the-hidden-dimensions-of-llm-alignment-a","title":"The Hidden Dimensions of LLM Alignment: A Multi-Dimensional Safety Analysis","date":"2025-02-13","arxiv_id":"2502.09674","repositories_listed":3,"syntology":null},{"url":"/paper/antidote-post-fine-tuning-safety-alignment","slug":"antidote-post-fine-tuning-safety-alignment","title":"Antidote: Post-fine-tuning Safety Alignment for Large Language Models against Harmful Fine-tuning","date":"2024-08-18","arxiv_id":"2408.09600","repositories_listed":3,"syntology":{"n":5,"n_ran":3,"n_constructed":0,"n_ran_checked":2,"n_instrument":1,"n_unverified":2,"n_honours":1,"n_violates":0,"n_no_contract":1,"n_pointer_only":0,"phrase":"3 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 1 honoured, 0 violated, 1 with no contract checked; 1 where Syntology's instrument failed) · 2 unverified","sample_list":"/paper/antidote-post-fine-tuning-safety-alignment#ran","syntology_url":"https://syntology.ai/paper/2408.09600","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2408.09600"}},"official":null}},{"url":"/paper/derail-yourself-multi-turn-llm-jailbreak","slug":"derail-yourself-multi-turn-llm-jailbreak","title":"Derail Yourself: Multi-turn LLM Jailbreak Attack through Self-discovered Clues","date":"2024-10-14","arxiv_id":"2410.10700","repositories_listed":2,"syntology":{"n":3,"n_ran":2,"n_constructed":0,"n_ran_checked":0,"n_instrument":2,"n_unverified":1,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":3,"phrase":"2 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 1 unverified","sample_list":"/paper/derail-yourself-multi-turn-llm-jailbreak#ran","syntology_url":"https://syntology.ai/paper/2410.10700","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2410.10700"}},"official":{"repos":["renqibing/actorattack"],"state":"official (archive's flag): 2 ran","n_ran":2,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":1,"ran_from_kinds":["official"]}}},{"url":"/paper/harmful-fine-tuning-attacks-and-defenses-for","slug":"harmful-fine-tuning-attacks-and-defenses-for","title":"Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey","date":"2024-09-26","arxiv_id":"2409.18169","repositories_listed":2,"syntology":{"n":9,"n_ran":6,"n_constructed":0,"n_ran_checked":3,"n_instrument":3,"n_unverified":3,"n_honours":2,"n_violates":1,"n_no_contract":0,"n_pointer_only":3,"phrase":"6 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 2 honoured, 1 violated, 0 with no contract checked; 3 where Syntology's instrument failed) · 3 unverified","sample_list":"/paper/harmful-fine-tuning-attacks-and-defenses-for#ran","syntology_url":"https://syntology.ai/paper/2409.18169","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2409.18169"}},"official":{"repos":["git-disl/awesome_llm-harmful-fine-tuning-papers"],"state":"community repositories only","n_ran":0,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":0,"ran_from_kinds":["listed"]}}},{"url":"/paper/exploring-safety-generalization-challenges-of","slug":"exploring-safety-generalization-challenges-of","title":"CodeAttack: Revealing Safety Generalization Challenges of Large Language Models via Code Completion","date":"2024-03-12","arxiv_id":"2403.07865","repositories_listed":2,"syntology":{"n":2,"n_ran":2,"n_constructed":0,"n_ran_checked":1,"n_instrument":1,"n_unverified":0,"n_honours":0,"n_violates":0,"n_no_contract":1,"n_pointer_only":0,"phrase":"2 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/exploring-safety-generalization-challenges-of#ran","syntology_url":"https://syntology.ai/paper/2403.07865","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2403.07865"}},"official":{"repos":["renqibing/CodeAttack"],"state":"official (archive's flag): 2 ran","n_ran":2,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/figstep-jailbreaking-large-vision-language","slug":"figstep-jailbreaking-large-vision-language","title":"FigStep: Jailbreaking Large Vision-Language Models via Typographic Visual Prompts","date":"2023-11-09","arxiv_id":"2311.05608","repositories_listed":2,"syntology":{"n":1,"n_ran":1,"n_constructed":0,"n_ran_checked":0,"n_instrument":1,"n_unverified":0,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":1,"phrase":"1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/figstep-jailbreaking-large-vision-language#ran","syntology_url":"https://syntology.ai/paper/2311.05608","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2311.05608"}},"official":{"repos":["thuccslab/figstep"],"state":"community repositories only","n_ran":0,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":0,"ran_from_kinds":["listed"]}}},{"url":"/paper/beyond-one-preference-for-all-multi-objective","slug":"beyond-one-preference-for-all-multi-objective","title":"Beyond One-Preference-Fits-All Alignment: Multi-Objective Direct Preference Optimization","date":"2023-10-05","arxiv_id":"2310.03708","repositories_listed":2,"syntology":null},{"url":"/paper/red-teaming-large-language-models-using-chain","slug":"red-teaming-large-language-models-using-chain","title":"Red-Teaming Large Language Models using Chain of Utterances for Safety-Alignment","date":"2023-08-18","arxiv_id":"2308.09662","repositories_listed":2,"syntology":null},{"url":"/paper/the-devil-behind-the-mask-an-emergent-safety","slug":"the-devil-behind-the-mask-an-emergent-safety","title":"The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs","date":"2025-07-15","arxiv_id":"2507.11097","repositories_listed":1,"syntology":{"n":17,"n_ran":16,"n_constructed":0,"n_ran_checked":3,"n_instrument":13,"n_unverified":1,"n_honours":2,"n_violates":0,"n_no_contract":1,"n_pointer_only":15,"phrase":"16 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 2 honoured, 0 violated, 1 with no contract checked; 13 where Syntology's instrument failed) · 1 unverified","sample_list":"/paper/the-devil-behind-the-mask-an-emergent-safety#ran","syntology_url":"https://syntology.ai/paper/2507.11097","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2507.11097"}},"official":{"repos":["zichenwen1/dija"],"state":"official (archive's flag): 16 ran","n_ran":16,"n_constructed":0,"n_ran_no_instrument_failure":3,"n_unverified":1,"ran_from_kinds":["official"]}}},{"url":"/paper/safe-pruning-lora-robust-distance-guided","slug":"safe-pruning-lora-robust-distance-guided","title":"Safe Pruning LoRA: Robust Distance-Guided Pruning for Safety Alignment in Adaptation of LLMs","date":"2025-06-21","arxiv_id":"2506.18931","repositories_listed":1,"syntology":{"n":8,"n_ran":6,"n_constructed":0,"n_ran_checked":3,"n_instrument":3,"n_unverified":2,"n_honours":0,"n_violates":0,"n_no_contract":3,"n_pointer_only":8,"phrase":"6 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 0 violated, 3 with no contract checked; 3 where Syntology's instrument failed) · 2 unverified","sample_list":"/paper/safe-pruning-lora-robust-distance-guided#ran","syntology_url":"https://syntology.ai/paper/2506.18931","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2506.18931"}},"official":{"repos":["aoshuang92/splora"],"state":"official (archive's flag): 6 ran","n_ran":6,"n_constructed":0,"n_ran_no_instrument_failure":3,"n_unverified":2,"ran_from_kinds":["official"]}}},{"url":"/paper/probe-before-you-talk-towards-black-box","slug":"probe-before-you-talk-towards-black-box","title":"Probe before You Talk: Towards Black-box Defense against Backdoor Unalignment for Large Language Models","date":"2025-06-19","arxiv_id":"2506.16447","repositories_listed":1,"syntology":null},{"url":"/paper/probing-the-robustness-of-large-language","slug":"probing-the-robustness-of-large-language","title":"Probing the Robustness of Large Language Models Safety to Latent Perturbations","date":"2025-06-19","arxiv_id":"2506.16078","repositories_listed":1,"syntology":null},{"url":"/paper/mitigating-safety-fallback-in-editing-based","slug":"mitigating-safety-fallback-in-editing-based","title":"Mitigating Safety Fallback in Editing-based Backdoor Injection on LLMs","date":"2025-06-16","arxiv_id":"2506.13285","repositories_listed":1,"syntology":null},{"url":"/paper/monitoring-decomposition-attacks-in-llms-with","slug":"monitoring-decomposition-attacks-in-llms-with","title":"Monitoring Decomposition Attacks in LLMs with Lightweight Sequential Monitors","date":"2025-06-12","arxiv_id":"2506.10949","repositories_listed":1,"syntology":null},{"url":"/paper/davsp-safety-alignment-for-large-vision","slug":"davsp-safety-alignment-for-large-vision","title":"DAVSP: Safety Alignment for Large Vision-Language Models via Deep Aligned Visual Safety Prompt","date":"2025-06-11","arxiv_id":"2506.09353","repositories_listed":1,"syntology":null},{"url":"/paper/chasing-moving-targets-with-online-self-play","slug":"chasing-moving-targets-with-online-self-play","title":"Chasing Moving Targets with Online Self-Play Reinforcement Learning for Safer Language Models","date":"2025-06-09","arxiv_id":"2506.07468","repositories_listed":1,"syntology":{"n":17,"n_ran":6,"n_constructed":0,"n_ran_checked":1,"n_instrument":5,"n_unverified":11,"n_honours":1,"n_violates":0,"n_no_contract":0,"n_pointer_only":1,"phrase":"6 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 1 honoured, 0 violated, 0 with no contract checked; 5 where Syntology's instrument failed) · 11 unverified","sample_list":"/paper/chasing-moving-targets-with-online-self-play#ran","syntology_url":"https://syntology.ai/paper/2506.07468","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2506.07468"}},"official":{"repos":["mickelliu/selfplay-redteaming"],"state":"official (archive's flag): 5 ran","n_ran":5,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":11,"ran_from_kinds":["official","unlocated"]}}},{"url":"/paper/rsafe-incentivizing-proactive-reasoning-to","slug":"rsafe-incentivizing-proactive-reasoning-to","title":"RSafe: Incentivizing proactive reasoning to build robust and adaptive LLM safeguards","date":"2025-06-09","arxiv_id":"2506.07736","repositories_listed":1,"syntology":null},{"url":"/paper/bitbypass-a-new-direction-in-jailbreaking","slug":"bitbypass-a-new-direction-in-jailbreaking","title":"BitBypass: A New Direction in Jailbreaking Aligned Large Language Models with Bitstream Camouflage","date":"2025-06-03","arxiv_id":"2506.02479","repositories_listed":1,"syntology":null},{"url":"/paper/diablo-diagonal-blocks-are-sufficient-for","slug":"diablo-diagonal-blocks-are-sufficient-for","title":"DiaBlo: Diagonal Blocks Are Sufficient For Finetuning","date":"2025-06-03","arxiv_id":"2506.03230","repositories_listed":1,"syntology":{"n":3,"n_ran":2,"n_constructed":1,"n_ran_checked":1,"n_instrument":1,"n_unverified":1,"n_honours":0,"n_violates":0,"n_no_contract":1,"n_pointer_only":3,"phrase":"2 ran (of which 1 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 1 where Syntology's instrument failed) · 1 unverified","sample_list":"/paper/diablo-diagonal-blocks-are-sufficient-for#ran","syntology_url":"https://syntology.ai/paper/2506.03230","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2506.03230"}},"official":{"repos":["ziyangjoy/diablo"],"state":"official (archive's flag): 2 ran","n_ran":2,"n_constructed":1,"n_ran_no_instrument_failure":1,"n_unverified":1,"ran_from_kinds":["official"]}}},{"url":"/paper/trident-enhancing-large-language-model-safety","slug":"trident-enhancing-large-language-model-safety","title":"TRIDENT: Enhancing Large Language Model Safety with Tri-Dimensional Diversified Red-Teaming Data Synthesis","date":"2025-05-30","arxiv_id":"2505.24672","repositories_listed":1,"syntology":null},{"url":"/paper/agentalign-navigating-safety-alignment-in-the","slug":"agentalign-navigating-safety-alignment-in-the","title":"AgentAlign: Navigating Safety Alignment in the Shift from Informative to Agentic Large Language Models","date":"2025-05-29","arxiv_id":"2505.23020","repositories_listed":1,"syntology":null},{"url":"/paper/overt-a-benchmark-for-over-refusal-evaluation","slug":"overt-a-benchmark-for-over-refusal-evaluation","title":"OVERT: A Benchmark for Over-Refusal Evaluation on Text-to-Image Models","date":"2025-05-27","arxiv_id":"2505.21347","repositories_listed":1,"syntology":{"n":1,"n_ran":0,"n_constructed":0,"n_ran_checked":0,"n_instrument":0,"n_unverified":1,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":0,"phrase":"0 ran · 1 unverified","sample_list":"/paper/overt-a-benchmark-for-over-refusal-evaluation#ran","syntology_url":"https://syntology.ai/paper/2505.21347","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2505.21347"}},"official":{"repos":["yixiao-huang/OVERT"],"state":"official: harvested, nothing ran","n_ran":0,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":1,"ran_from_kinds":[]}}},{"url":"/paper/beyond-safe-answers-a-benchmark-for","slug":"beyond-safe-answers-a-benchmark-for","title":"Beyond Safe Answers: A Benchmark for Evaluating True Risk Awareness in Large Reasoning Models","date":"2025-05-26","arxiv_id":"2505.19690","repositories_listed":1,"syntology":null},{"url":"/paper/lifelong-safety-alignment-for-language-models","slug":"lifelong-safety-alignment-for-language-models","title":"Lifelong Safety Alignment for Language Models","date":"2025-05-26","arxiv_id":"2505.20259","repositories_listed":1,"syntology":null},{"url":"/paper/vscbench-bridging-the-gap-in-vision-language","slug":"vscbench-bridging-the-gap-in-vision-language","title":"VSCBench: Bridging the Gap in Vision-Language Model Safety Calibration","date":"2025-05-26","arxiv_id":"2505.20362","repositories_listed":1,"syntology":null},{"url":"/paper/one-model-transfer-to-all-on-robust-jailbreak","slug":"one-model-transfer-to-all-on-robust-jailbreak","title":"One Model Transfer to All: On Robust Jailbreak Prompts Generation against LLMs","date":"2025-05-23","arxiv_id":"2505.17598","repositories_listed":1,"syntology":null},{"url":"/paper/duffin-a-dual-level-fingerprinting-framework","slug":"duffin-a-dual-level-fingerprinting-framework","title":"DuFFin: A Dual-Level Fingerprinting Framework for LLMs IP Protection","date":"2025-05-22","arxiv_id":"2505.16530","repositories_listed":1,"syntology":null},{"url":"/paper/mitigating-fine-tuning-risks-in-llms-via","slug":"mitigating-fine-tuning-risks-in-llms-via","title":"Mitigating Fine-tuning Risks in LLMs via Safety-Aware Probing Optimization","date":"2025-05-22","arxiv_id":"2505.16737","repositories_listed":1,"syntology":null},{"url":"/paper/mpo-multilingual-safety-alignment-via-reward","slug":"mpo-multilingual-safety-alignment-via-reward","title":"MPO: Multilingual Safety Alignment via Reward Gap Optimization","date":"2025-05-22","arxiv_id":"2505.16869","repositories_listed":1,"syntology":{"n":3,"n_ran":3,"n_constructed":0,"n_ran_checked":1,"n_instrument":2,"n_unverified":0,"n_honours":0,"n_violates":0,"n_no_contract":1,"n_pointer_only":0,"phrase":"3 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/mpo-multilingual-safety-alignment-via-reward#ran","syntology_url":"https://syntology.ai/paper/2505.16869","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2505.16869"}},"official":{"repos":["circle-hit/mpo"],"state":"official (archive's flag): 3 ran","n_ran":3,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/mtsa-multi-turn-safety-alignment-for-llms","slug":"mtsa-multi-turn-safety-alignment-for-llms","title":"MTSA: Multi-turn Safety Alignment for LLMs through Multi-round Red-teaming","date":"2025-05-22","arxiv_id":"2505.17147","repositories_listed":1,"syntology":null},{"url":"/paper/keep-security-benchmarking-security-policy","slug":"keep-security-benchmarking-security-policy","title":"Keep Security! Benchmarking Security Policy Preservation in Large Language Model Contexts Against Indirect Attacks in Question Answering","date":"2025-05-21","arxiv_id":"2505.15805","repositories_listed":1,"syntology":null},{"url":"/paper/linear-control-of-test-awareness-reveals","slug":"linear-control-of-test-awareness-reveals","title":"Linear Control of Test Awareness Reveals Differential Compliance in Reasoning Models","date":"2025-05-20","arxiv_id":"2505.14617","repositories_listed":1,"syntology":{"n":4,"n_ran":1,"n_constructed":0,"n_ran_checked":0,"n_instrument":1,"n_unverified":3,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":0,"phrase":"1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 3 unverified","sample_list":"/paper/linear-control-of-test-awareness-reveals#ran","syntology_url":"https://syntology.ai/paper/2505.14617","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2505.14617"}},"official":{"repos":["microsoft/test_awareness_steering"],"state":"official (archive's flag): 1 ran","n_ran":1,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":3,"ran_from_kinds":["official"]}}},{"url":"/paper/pandaguard-systematic-evaluation-of-llm","slug":"pandaguard-systematic-evaluation-of-llm","title":"PandaGuard: Systematic Evaluation of LLM Safety against Jailbreaking Attacks","date":"2025-05-20","arxiv_id":"2505.13862","repositories_listed":1,"syntology":null},{"url":"/paper/safety-subspaces-are-not-distinct-a-fine","slug":"safety-subspaces-are-not-distinct-a-fine","title":"Safety Subspaces are Not Distinct: A Fine-Tuning Case Study","date":"2025-05-20","arxiv_id":"2505.14185","repositories_listed":1,"syntology":{"n":4,"n_ran":3,"n_constructed":0,"n_ran_checked":1,"n_instrument":2,"n_unverified":1,"n_honours":0,"n_violates":0,"n_no_contract":1,"n_pointer_only":4,"phrase":"3 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 2 where Syntology's instrument failed) · 1 unverified","sample_list":"/paper/safety-subspaces-are-not-distinct-a-fine#ran","syntology_url":"https://syntology.ai/paper/2505.14185","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2505.14185"}},"official":{"repos":["cert-lab/safety-subspaces"],"state":"official (archive's flag): 3 ran","n_ran":3,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":1,"ran_from_kinds":["official"]}}},{"url":"/paper/2505-11063","slug":"2505-11063","title":"Think Twice Before You Act: Enhancing Agent Behavioral Safety with Thought Correction","date":"2025-05-16","arxiv_id":"2505.11063","repositories_listed":1,"syntology":null},{"url":"/paper/benign-samples-matter-fine-tuning-on-outlier","slug":"benign-samples-matter-fine-tuning-on-outlier","title":"Benign Samples Matter! Fine-tuning On Outlier Benign Samples Severely Breaks Safety","date":"2025-05-11","arxiv_id":"2505.06843","repositories_listed":1,"syntology":{"n":1,"n_ran":1,"n_constructed":0,"n_ran_checked":1,"n_instrument":0,"n_unverified":0,"n_honours":1,"n_violates":0,"n_no_contract":0,"n_pointer_only":1,"phrase":"1 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 1 honoured, 0 violated, 0 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/benign-samples-matter-fine-tuning-on-outlier#ran","syntology_url":"https://syntology.ai/paper/2505.06843","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2505.06843"}},"official":{"repos":["guanzihan/benign-samples-matter"],"state":"official (archive's flag): 1 ran","n_ran":1,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/think-in-safety-unveiling-and-mitigating","slug":"think-in-safety-unveiling-and-mitigating","title":"Think in Safety: Unveiling and Mitigating Safety Alignment Collapse in Multimodal Large Reasoning Model","date":"2025-05-10","arxiv_id":"2505.06538","repositories_listed":1,"syntology":null},{"url":"/paper/texttt-sage-a-generic-framework-for-llm","slug":"texttt-sage-a-generic-framework-for-llm","title":"$\\texttt{SAGE}$: A Generic Framework for LLM Safety Evaluation","date":"2025-04-28","arxiv_id":"2504.19674","repositories_listed":1,"syntology":null},{"url":"/paper/dream-disentangling-risks-to-enhance-safety","slug":"dream-disentangling-risks-to-enhance-safety","title":"DREAM: Disentangling Risks to Enhance Safety Alignment in Multimodal Large Language Models","date":"2025-04-25","arxiv_id":"2504.18053","repositories_listed":1,"syntology":null},{"url":"/paper/adasteer-your-aligned-llm-is-inherently-an","slug":"adasteer-your-aligned-llm-is-inherently-an","title":"AdaSteer: Your Aligned LLM is Inherently an Adaptive Jailbreak Defender","date":"2025-04-13","arxiv_id":"2504.09466","repositories_listed":1,"syntology":null},{"url":"/paper/lori-reducing-cross-task-interference-in","slug":"lori-reducing-cross-task-interference-in","title":"LoRI: Reducing Cross-Task Interference in Multi-Task Low-Rank Adaptation","date":"2025-04-10","arxiv_id":"2504.07448","repositories_listed":1,"syntology":null},{"url":"/paper/sudo-rm-rf-agentic-security","slug":"sudo-rm-rf-agentic-security","title":"sudo rm -rf agentic_security","date":"2025-03-26","arxiv_id":"2503.20279","repositories_listed":1,"syntology":{"n":2,"n_ran":2,"n_constructed":0,"n_ran_checked":0,"n_instrument":2,"n_unverified":0,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":2,"phrase":"2 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/sudo-rm-rf-agentic-security#ran","syntology_url":"https://syntology.ai/paper/2503.20279","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2503.20279"}},"official":{"repos":["AIM-Intelligence/SUDO"],"state":"official (archive's flag): 2 ran","n_ran":2,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/vpo-aligning-text-to-video-generation-models","slug":"vpo-aligning-text-to-video-generation-models","title":"VPO: Aligning Text-to-Video Generation Models with Prompt Optimization","date":"2025-03-26","arxiv_id":"2503.20491","repositories_listed":1,"syntology":null},{"url":"/paper/lookahead-tuning-safer-language-models-via","slug":"lookahead-tuning-safer-language-models-via","title":"LookAhead Tuning: Safer Language Models via Partial Answer Previews","date":"2025-03-24","arxiv_id":"2503.19041","repositories_listed":1,"syntology":null},{"url":"/paper/safemerge-preserving-safety-alignment-in-fine","slug":"safemerge-preserving-safety-alignment-in-fine","title":"SafeMERGE: Preserving Safety Alignment in Fine-Tuned Large Language Models via Selective Layer-Wise Model Merging","date":"2025-03-21","arxiv_id":"2503.17239","repositories_listed":1,"syntology":{"n":1,"n_ran":1,"n_constructed":0,"n_ran_checked":0,"n_instrument":1,"n_unverified":0,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":1,"phrase":"1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/safemerge-preserving-safety-alignment-in-fine#ran","syntology_url":"https://syntology.ai/paper/2503.17239","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2503.17239"}},"official":{"repos":["aladinD/SafeMERGE"],"state":"official (archive's flag): 1 ran","n_ran":1,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/improving-llm-safety-alignment-with-dual","slug":"improving-llm-safety-alignment-with-dual","title":"Improving LLM Safety Alignment with Dual-Objective Optimization","date":"2025-03-05","arxiv_id":"2503.03710","repositories_listed":1,"syntology":{"n":5,"n_ran":5,"n_constructed":0,"n_ran_checked":2,"n_instrument":3,"n_unverified":0,"n_honours":1,"n_violates":1,"n_no_contract":0,"n_pointer_only":5,"phrase":"5 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 1 honoured, 1 violated, 0 with no contract checked; 3 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/improving-llm-safety-alignment-with-dual#ran","syntology_url":"https://syntology.ai/paper/2503.03710","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2503.03710"}},"official":{"repos":["wicai24/door-alignment"],"state":"official (archive's flag): 5 ran","n_ran":5,"n_constructed":0,"n_ran_no_instrument_failure":2,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/2503-00555","slug":"2503-00555","title":"Safety Tax: Safety Alignment Makes Your Large Reasoning Models Less Reasonable","date":"2025-03-01","arxiv_id":"2503.00555","repositories_listed":1,"syntology":{"n":5,"n_ran":4,"n_constructed":0,"n_ran_checked":2,"n_instrument":2,"n_unverified":1,"n_honours":0,"n_violates":1,"n_no_contract":1,"n_pointer_only":1,"phrase":"4 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 1 violated, 1 with no contract checked; 2 where Syntology's instrument failed) · 1 unverified","sample_list":"/paper/2503-00555#ran","syntology_url":"https://syntology.ai/paper/2503.00555","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2503.00555"}},"official":{"repos":["git-disl/safety-tax"],"state":"official (archive's flag): 4 ran","n_ran":4,"n_constructed":0,"n_ran_no_instrument_failure":2,"n_unverified":1,"ran_from_kinds":["official"]}}},{"url":"/paper/steering-dialogue-dynamics-for-robustness","slug":"steering-dialogue-dynamics-for-robustness","title":"Steering Dialogue Dynamics for Robustness against Multi-turn Jailbreaking Attacks","date":"2025-02-28","arxiv_id":"2503.00187","repositories_listed":1,"syntology":{"n":2,"n_ran":1,"n_constructed":0,"n_ran_checked":0,"n_instrument":1,"n_unverified":1,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":0,"phrase":"1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 1 unverified","sample_list":"/paper/steering-dialogue-dynamics-for-robustness#ran","syntology_url":"https://syntology.ai/paper/2503.00187","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2503.00187"}},"official":{"repos":["hanjianghu/nbf-llm"],"state":"official (archive's flag): 1 ran","n_ran":1,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":1,"ran_from_kinds":["official"]}}},{"url":"/paper/exploiting-prefix-tree-in-structured-output","slug":"exploiting-prefix-tree-in-structured-output","title":"Exploiting Prefix-Tree in Structured Output Interfaces for Enhancing Jailbreak Attacking","date":"2025-02-19","arxiv_id":"2502.13527","repositories_listed":1,"syntology":null},{"url":"/paper/sea-low-resource-safety-alignment-for","slug":"sea-low-resource-safety-alignment-for","title":"SEA: Low-Resource Safety Alignment for Multimodal Large Language Models via Synthetic Embeddings","date":"2025-02-18","arxiv_id":"2502.12562","repositories_listed":1,"syntology":null},{"url":"/paper/structtransform-a-scalable-attack-surface-for","slug":"structtransform-a-scalable-attack-surface-for","title":"StructTransform: A Scalable Attack Surface for Safety-Aligned Large Language Models","date":"2025-02-17","arxiv_id":"2502.11853","repositories_listed":1,"syntology":null},{"url":"/paper/reasoning-augmented-conversation-for-multi","slug":"reasoning-augmented-conversation-for-multi","title":"Reasoning-Augmented Conversation for Multi-Turn Jailbreak Attacks on Large Language Models","date":"2025-02-16","arxiv_id":"2502.11054","repositories_listed":1,"syntology":null},{"url":"/paper/soteria-language-specific-functional","slug":"soteria-language-specific-functional","title":"Soteria: Language-Specific Functional Parameter Steering for Multilingual Safety Alignment","date":"2025-02-16","arxiv_id":"2502.11244","repositories_listed":1,"syntology":null},{"url":"/paper/x-boundary-establishing-exact-safety-boundary","slug":"x-boundary-establishing-exact-safety-boundary","title":"X-Boundary: Establishing Exact Safety Boundary to Shield LLMs from Multi-Turn Jailbreaks without Compromising Usability","date":"2025-02-14","arxiv_id":"2502.09990","repositories_listed":1,"syntology":{"n":5,"n_ran":5,"n_constructed":0,"n_ran_checked":1,"n_instrument":4,"n_unverified":0,"n_honours":1,"n_violates":0,"n_no_contract":0,"n_pointer_only":5,"phrase":"5 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 1 honoured, 0 violated, 0 with no contract checked; 4 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/x-boundary-establishing-exact-safety-boundary#ran","syntology_url":"https://syntology.ai/paper/2502.09990","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2502.09990"}},"official":{"repos":["ai45lab/x-boundary"],"state":"official (archive's flag): 5 ran","n_ran":5,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/making-them-a-malicious-database-exploiting","slug":"making-them-a-malicious-database-exploiting","title":"QueryAttack: Jailbreaking Aligned Large Language Models Using Structured Non-natural Query Language","date":"2025-02-13","arxiv_id":"2502.09723","repositories_listed":1,"syntology":{"n":3,"n_ran":3,"n_constructed":0,"n_ran_checked":1,"n_instrument":2,"n_unverified":0,"n_honours":1,"n_violates":0,"n_no_contract":0,"n_pointer_only":0,"phrase":"3 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 1 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/making-them-a-malicious-database-exploiting#ran","syntology_url":"https://syntology.ai/paper/2502.09723","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2502.09723"}},"official":{"repos":["horizonsinzqs/queryattack"],"state":"official (archive's flag): 3 ran","n_ran":3,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/speak-easy-eliciting-harmful-jailbreaks-from","slug":"speak-easy-eliciting-harmful-jailbreaks-from","title":"Speak Easy: Eliciting Harmful Jailbreaks from LLMs with Simple Interactions","date":"2025-02-06","arxiv_id":"2502.04322","repositories_listed":1,"syntology":{"n":2,"n_ran":2,"n_constructed":0,"n_ran_checked":2,"n_instrument":0,"n_unverified":0,"n_honours":0,"n_violates":0,"n_no_contract":2,"n_pointer_only":0,"phrase":"2 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 0 violated, 2 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/speak-easy-eliciting-harmful-jailbreaks-from#ran","syntology_url":"https://syntology.ai/paper/2502.04322","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2502.04322"}},"official":{"repos":["yiksiu-chan/SpeakEasy"],"state":"official (archive's flag): 2 ran","n_ran":2,"n_constructed":0,"n_ran_no_instrument_failure":2,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/pandas-improving-many-shot-jailbreaking-via","slug":"pandas-improving-many-shot-jailbreaking-via","title":"PANDAS: Improving Many-shot Jailbreaking via Positive Affirmation, Negative Demonstration, and Adaptive Sampling","date":"2025-02-04","arxiv_id":"2502.01925","repositories_listed":1,"syntology":{"n":1,"n_ran":1,"n_constructed":0,"n_ran_checked":0,"n_instrument":1,"n_unverified":0,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":1,"phrase":"1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/pandas-improving-many-shot-jailbreaking-via#ran","syntology_url":"https://syntology.ai/paper/2502.01925","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2502.01925"}},"official":{"repos":["averyma/pandas"],"state":"official (archive's flag): 1 ran","n_ran":1,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/stair-improving-safety-alignment-with","slug":"stair-improving-safety-alignment-with","title":"STAIR: Improving Safety Alignment with Introspective Reasoning","date":"2025-02-04","arxiv_id":"2502.02384","repositories_listed":1,"syntology":{"n":6,"n_ran":6,"n_constructed":0,"n_ran_checked":5,"n_instrument":1,"n_unverified":0,"n_honours":0,"n_violates":0,"n_no_contract":5,"n_pointer_only":1,"phrase":"6 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 0 honoured, 0 violated, 5 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/stair-improving-safety-alignment-with#ran","syntology_url":"https://syntology.ai/paper/2502.02384","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2502.02384"}},"official":{"repos":["thu-ml/stair"],"state":"official (archive's flag): 6 ran","n_ran":6,"n_constructed":0,"n_ran_no_instrument_failure":5,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/llm-safety-alignment-is-divergence-estimation","slug":"llm-safety-alignment-is-divergence-estimation","title":"LLM Safety Alignment is Divergence Estimation in Disguise","date":"2025-02-02","arxiv_id":"2502.00657","repositories_listed":1,"syntology":{"n":17,"n_ran":9,"n_constructed":0,"n_ran_checked":9,"n_instrument":0,"n_unverified":8,"n_honours":0,"n_violates":0,"n_no_contract":9,"n_pointer_only":2,"phrase":"9 ran (of which 0 constructed an object rather than computing a result; 9 with no instrument failure: 0 honoured, 0 violated, 9 with no contract checked; 0 where Syntology's instrument failed) · 8 unverified","sample_list":"/paper/llm-safety-alignment-is-divergence-estimation#ran","syntology_url":"https://syntology.ai/paper/2502.00657","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2502.00657"}},"official":{"repos":["rhaldarpurdue/kldo"],"state":"official (archive's flag): 9 ran","n_ran":9,"n_constructed":0,"n_ran_no_instrument_failure":9,"n_unverified":8,"ran_from_kinds":["official"]}}},{"url":"/paper/panacea-mitigating-harmful-fine-tuning-for","slug":"panacea-mitigating-harmful-fine-tuning-for","title":"Panacea: Mitigating Harmful Fine-tuning for Large Language Models via Post-fine-tuning Perturbation","date":"2025-01-30","arxiv_id":"2501.18100","repositories_listed":1,"syntology":null},{"url":"/paper/virus-harmful-fine-tuning-attack-for-large","slug":"virus-harmful-fine-tuning-attack-for-large","title":"Virus: Harmful Fine-tuning Attack for Large Language Models Bypassing Guardrail Moderation","date":"2025-01-29","arxiv_id":"2501.17433","repositories_listed":1,"syntology":{"n":11,"n_ran":8,"n_constructed":0,"n_ran_checked":5,"n_instrument":3,"n_unverified":3,"n_honours":2,"n_violates":0,"n_no_contract":3,"n_pointer_only":3,"phrase":"8 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 2 honoured, 0 violated, 3 with no contract checked; 3 where Syntology's instrument failed) · 3 unverified","sample_list":"/paper/virus-harmful-fine-tuning-attack-for-large#ran","syntology_url":"https://syntology.ai/paper/2501.17433","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2501.17433"}},"official":{"repos":["git-disl/virus"],"state":"official (archive's flag): 8 ran","n_ran":8,"n_constructed":0,"n_ran_no_instrument_failure":5,"n_unverified":3,"ran_from_kinds":["official"]}}},{"url":"/paper/xjailbreak-representation-space-guided","slug":"xjailbreak-representation-space-guided","title":"xJailbreak: Representation Space Guided Reinforcement Learning for Interpretable LLM Jailbreaking","date":"2025-01-28","arxiv_id":"2501.16727","repositories_listed":1,"syntology":null},{"url":"/paper/latent-space-adversarial-training-with-post","slug":"latent-space-adversarial-training-with-post","title":"Latent-space adversarial training with post-aware calibration for defending large language models against jailbreak attacks","date":"2025-01-18","arxiv_id":"2501.10639","repositories_listed":1,"syntology":null},{"url":"/paper/sata-a-paradigm-for-llm-jailbreak-via-simple","slug":"sata-a-paradigm-for-llm-jailbreak-via-simple","title":"SATA: A Paradigm for LLM Jailbreak via Simple Assistive Task Linkage","date":"2024-12-19","arxiv_id":"2412.15289","repositories_listed":1,"syntology":null},{"url":"/paper/autonomous-microscopy-experiments-through","slug":"autonomous-microscopy-experiments-through","title":"Autonomous Microscopy Experiments through Large Language Model Agents","date":"2024-12-18","arxiv_id":"2501.10385","repositories_listed":1,"syntology":null},{"url":"/paper/separate-the-wheat-from-the-chaff-a-post-hoc","slug":"separate-the-wheat-from-the-chaff-a-post-hoc","title":"Separate the Wheat from the Chaff: A Post-Hoc Approach to Safety Re-Alignment for Fine-Tuned Language Models","date":"2024-12-15","arxiv_id":"2412.11041","repositories_listed":1,"syntology":{"n":2,"n_ran":1,"n_constructed":0,"n_ran_checked":0,"n_instrument":1,"n_unverified":1,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":2,"phrase":"1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 1 unverified","sample_list":"/paper/separate-the-wheat-from-the-chaff-a-post-hoc#ran","syntology_url":"https://syntology.ai/paper/2412.11041","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2412.11041"}},"official":{"repos":["pikepokenew/IRR"],"state":"official (archive's flag): 1 ran","n_ran":1,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":1,"ran_from_kinds":["official"]}}},{"url":"/paper/safeworld-geo-diverse-safety-alignment","slug":"safeworld-geo-diverse-safety-alignment","title":"SafeWorld: Geo-Diverse Safety Alignment","date":"2024-12-09","arxiv_id":"2412.06483","repositories_listed":1,"syntology":null},{"url":"/paper/privagent-agentic-based-red-teaming-for-llm","slug":"privagent-agentic-based-red-teaming-for-llm","title":"PrivAgent: Agentic-based Red-teaming for LLM Privacy Leakage","date":"2024-12-07","arxiv_id":"2412.05734","repositories_listed":1,"syntology":null},{"url":"/paper/immune-improving-safety-against-jailbreaks-in","slug":"immune-improving-safety-against-jailbreaks-in","title":"Immune: Improving Safety Against Jailbreaks in Multi-modal LLMs via Inference-Time Alignment","date":"2024-11-27","arxiv_id":"2411.18688","repositories_listed":1,"syntology":{"n":2,"n_ran":1,"n_constructed":0,"n_ran_checked":0,"n_instrument":1,"n_unverified":1,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":2,"phrase":"1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 1 unverified","sample_list":"/paper/immune-improving-safety-against-jailbreaks-in#ran","syntology_url":"https://syntology.ai/paper/2411.18688","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2411.18688"}},"official":null}},{"url":"/paper/don-t-command-cultivate-an-exploratory-study","slug":"don-t-command-cultivate-an-exploratory-study","title":"Don't Command, Cultivate: An Exploratory Study of System-2 Alignment","date":"2024-11-26","arxiv_id":"2411.17075","repositories_listed":1,"syntology":null},{"url":"/paper/stochastic-monkeys-at-play-random","slug":"stochastic-monkeys-at-play-random","title":"Stochastic Monkeys at Play: Random Augmentations Cheaply Break LLM Safety Alignment","date":"2024-11-05","arxiv_id":"2411.02785","repositories_listed":1,"syntology":null},{"url":"/paper/audio-is-the-achilles-heel-red-teaming-audio","slug":"audio-is-the-achilles-heel-red-teaming-audio","title":"Audio Is the Achilles' Heel: Red Teaming Audio Large Multimodal Models","date":"2024-10-31","arxiv_id":"2410.23861","repositories_listed":1,"syntology":null},{"url":"/paper/sg-bench-evaluating-llm-safety-generalization","slug":"sg-bench-evaluating-llm-safety-generalization","title":"SG-Bench: Evaluating LLM Safety Generalization Across Diverse Tasks and Prompt Types","date":"2024-10-29","arxiv_id":"2410.21965","repositories_listed":1,"syntology":{"n":1,"n_ran":1,"n_constructed":0,"n_ran_checked":0,"n_instrument":1,"n_unverified":0,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":1,"phrase":"1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/sg-bench-evaluating-llm-safety-generalization#ran","syntology_url":"https://syntology.ai/paper/2410.21965","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2410.21965"}},"official":{"repos":["MurrayTom/SG-Bench"],"state":"official (archive's flag): 1 ran","n_ran":1,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/enhancing-safety-in-reinforcement-learning","slug":"enhancing-safety-in-reinforcement-learning","title":"Enhancing Safety in Reinforcement Learning with Human Feedback via Rectified Policy Optimization","date":"2024-10-25","arxiv_id":"2410.19933","repositories_listed":1,"syntology":{"n":1,"n_ran":1,"n_constructed":0,"n_ran_checked":1,"n_instrument":0,"n_unverified":0,"n_honours":0,"n_violates":1,"n_no_contract":0,"n_pointer_only":0,"phrase":"1 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 1 violated, 0 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/enhancing-safety-in-reinforcement-learning#ran","syntology_url":"https://syntology.ai/paper/2410.19933","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2410.19933"}},"official":{"repos":["pxywatermoon/repo"],"state":"official (archive's flag): 1 ran","n_ran":1,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/iterative-self-tuning-llms-for-enhanced","slug":"iterative-self-tuning-llms-for-enhanced","title":"Iterative Self-Tuning LLMs for Enhanced Jailbreaking Capabilities","date":"2024-10-24","arxiv_id":"2410.18469","repositories_listed":1,"syntology":null},{"url":"/paper/bayesian-scaling-laws-for-in-context-learning","slug":"bayesian-scaling-laws-for-in-context-learning","title":"Bayesian scaling laws for in-context learning","date":"2024-10-21","arxiv_id":"2410.16531","repositories_listed":1,"syntology":{"n":3,"n_ran":2,"n_constructed":0,"n_ran_checked":0,"n_instrument":2,"n_unverified":1,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":3,"phrase":"2 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 1 unverified","sample_list":"/paper/bayesian-scaling-laws-for-in-context-learning#ran","syntology_url":"https://syntology.ai/paper/2410.16531","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2410.16531"}},"official":{"repos":["aryamanarora/bayesian-laws-icl"],"state":"official (archive's flag): 2 ran","n_ran":2,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":1,"ran_from_kinds":["official"]}}},{"url":"/paper/a-common-pitfall-of-margin-based-language","slug":"a-common-pitfall-of-margin-based-language","title":"A Common Pitfall of Margin-based Language Model Alignment: Gradient Entanglement","date":"2024-10-17","arxiv_id":"2410.13828","repositories_listed":1,"syntology":{"n":5,"n_ran":4,"n_constructed":0,"n_ran_checked":1,"n_instrument":3,"n_unverified":1,"n_honours":0,"n_violates":0,"n_no_contract":1,"n_pointer_only":5,"phrase":"4 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 3 where Syntology's instrument failed) · 1 unverified","sample_list":"/paper/a-common-pitfall-of-margin-based-language#ran","syntology_url":"https://syntology.ai/paper/2410.13828","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2410.13828"}},"official":{"repos":["humainlab/understand_marginpo"],"state":"official (archive's flag): 4 ran","n_ran":4,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":1,"ran_from_kinds":["official"]}}},{"url":"/paper/do-llms-have-political-correctness-analyzing-1","slug":"do-llms-have-political-correctness-analyzing-1","title":"BiasJailbreak:Analyzing Ethical Biases and Jailbreak Vulnerabilities in Large Language Models","date":"2024-10-17","arxiv_id":"2410.13334","repositories_listed":1,"syntology":null},{"url":"/paper/locking-down-the-finetuned-llms-safety","slug":"locking-down-the-finetuned-llms-safety","title":"Locking Down the Finetuned LLMs Safety","date":"2024-10-14","arxiv_id":"2410.10343","repositories_listed":1,"syntology":{"n":21,"n_ran":12,"n_constructed":0,"n_ran_checked":3,"n_instrument":9,"n_unverified":9,"n_honours":0,"n_violates":0,"n_no_contract":3,"n_pointer_only":21,"phrase":"12 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 0 violated, 3 with no contract checked; 9 where Syntology's instrument failed) · 9 unverified","sample_list":"/paper/locking-down-the-finetuned-llms-safety#ran","syntology_url":"https://syntology.ai/paper/2410.10343","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2410.10343"}},"official":{"repos":["zhu-minjun/safetylock"],"state":"official (archive's flag): 12 ran","n_ran":12,"n_constructed":0,"n_ran_no_instrument_failure":3,"n_unverified":9,"ran_from_kinds":["official"]}}},{"url":"/paper/targeted-vaccine-safety-alignment-for-large","slug":"targeted-vaccine-safety-alignment-for-large","title":"Targeted Vaccine: Safety Alignment for Large Language Models against Harmful Fine-Tuning via Layer-wise Perturbation","date":"2024-10-13","arxiv_id":"2410.09760","repositories_listed":1,"syntology":{"n":12,"n_ran":12,"n_constructed":0,"n_ran_checked":1,"n_instrument":11,"n_unverified":0,"n_honours":1,"n_violates":0,"n_no_contract":0,"n_pointer_only":3,"phrase":"12 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 1 honoured, 0 violated, 0 with no contract checked; 11 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/targeted-vaccine-safety-alignment-for-large#ran","syntology_url":"https://syntology.ai/paper/2410.09760","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2410.09760"}},"official":{"repos":["lslland/t-vaccine"],"state":"official (archive's flag): 12 ran","n_ran":12,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/attngcg-enhancing-jailbreaking-attacks-on","slug":"attngcg-enhancing-jailbreaking-attacks-on","title":"AttnGCG: Enhancing Jailbreaking Attacks on LLMs with Attention Manipulation","date":"2024-10-11","arxiv_id":"2410.09040","repositories_listed":1,"syntology":null},{"url":"/paper/can-a-large-language-model-be-a-gaslighter","slug":"can-a-large-language-model-be-a-gaslighter","title":"Can a large language model be a gaslighter?","date":"2024-10-11","arxiv_id":"2410.09181","repositories_listed":1,"syntology":null},{"url":"/paper/data-advisor-dynamic-data-curation-for-safety","slug":"data-advisor-dynamic-data-curation-for-safety","title":"Data Advisor: Dynamic Data Curation for Safety Alignment of Large Language Models","date":"2024-10-07","arxiv_id":"2410.05269","repositories_listed":1,"syntology":{"n":2,"n_ran":2,"n_constructed":0,"n_ran_checked":0,"n_instrument":2,"n_unverified":0,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":2,"phrase":"2 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/data-advisor-dynamic-data-curation-for-safety#ran","syntology_url":"https://syntology.ai/paper/2410.05269","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2410.05269"}},"official":{"repos":["feiwang96/Data-Advisor"],"state":"official (archive's flag): 2 ran","n_ran":2,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/adversarial-suffixes-may-be-features-too","slug":"adversarial-suffixes-may-be-features-too","title":"Unleashing the Unseen: Harnessing Benign Datasets for Jailbreaking Large Language Models","date":"2024-10-01","arxiv_id":"2410.00451","repositories_listed":1,"syntology":null},{"url":"/paper/overriding-safety-protections-of-open-source","slug":"overriding-safety-protections-of-open-source","title":"Overriding Safety protections of Open-source Models","date":"2024-09-28","arxiv_id":"2409.19476","repositories_listed":1,"syntology":null},{"url":"/paper/nothing-in-excess-mitigating-the-exaggerated","slug":"nothing-in-excess-mitigating-the-exaggerated","title":"SCANS: Mitigating the Exaggerated Safety for LLMs via Safety-Conscious Activation Steering","date":"2024-08-21","arxiv_id":"2408.11491","repositories_listed":1,"syntology":{"n":2,"n_ran":2,"n_constructed":0,"n_ran_checked":2,"n_instrument":0,"n_unverified":0,"n_honours":2,"n_violates":0,"n_no_contract":0,"n_pointer_only":2,"phrase":"2 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 2 honoured, 0 violated, 0 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/nothing-in-excess-mitigating-the-exaggerated#ran","syntology_url":"https://syntology.ai/paper/2408.11491","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2408.11491"}},"official":{"repos":["zouyingcao/SCANS"],"state":"official (archive's flag): 2 ran","n_ran":2,"n_constructed":0,"n_ran_no_instrument_failure":2,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/unlocking-adversarial-suffix-optimization","slug":"unlocking-adversarial-suffix-optimization","title":"Unlocking Adversarial Suffix Optimization Without Affirmative Phrases: Efficient Black-box Jailbreaking via LLM as Optimizer","date":"2024-08-21","arxiv_id":"2408.11313","repositories_listed":1,"syntology":{"n":2,"n_ran":2,"n_constructed":0,"n_ran_checked":1,"n_instrument":1,"n_unverified":0,"n_honours":0,"n_violates":1,"n_no_contract":0,"n_pointer_only":2,"phrase":"2 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 1 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/unlocking-adversarial-suffix-optimization#ran","syntology_url":"https://syntology.ai/paper/2408.11313","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2408.11313"}},"official":{"repos":["lenijwp/eclipse"],"state":"official (archive's flag): 2 ran","n_ran":2,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/ferret-faster-and-effective-automated-red","slug":"ferret-faster-and-effective-automated-red","title":"Ferret: Faster and Effective Automated Red Teaming with Reward-Based Scoring Technique","date":"2024-08-20","arxiv_id":"2408.10701","repositories_listed":1,"syntology":null},{"url":"/paper/alignment-enhanced-decoding-defending-via","slug":"alignment-enhanced-decoding-defending-via","title":"Alignment-Enhanced Decoding:Defending via Token-Level Adaptive Refining of Probability Distributions","date":"2024-08-14","arxiv_id":"2408.07663","repositories_listed":1,"syntology":null},{"url":"/paper/2407-21659","slug":"2407-21659","title":"Cross-modality Information Check for Detecting Jailbreaking in Multimodal Large Language Models","date":"2024-07-31","arxiv_id":"2407.21659","repositories_listed":1,"syntology":{"n":8,"n_ran":7,"n_constructed":0,"n_ran_checked":1,"n_instrument":6,"n_unverified":1,"n_honours":0,"n_violates":0,"n_no_contract":1,"n_pointer_only":0,"phrase":"7 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 6 where Syntology's instrument failed) · 1 unverified","sample_list":"/paper/2407-21659#ran","syntology_url":"https://syntology.ai/paper/2407.21659","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2407.21659"}},"official":{"repos":["pandragonxiii/cider"],"state":"official (archive's flag): 7 ran","n_ran":7,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":1,"ran_from_kinds":["official"]}}},{"url":"/paper/can-editing-llms-inject-harm","slug":"can-editing-llms-inject-harm","title":"Can Editing LLMs Inject Harm?","date":"2024-07-29","arxiv_id":"2407.20224","repositories_listed":1,"syntology":{"n":13,"n_ran":10,"n_constructed":0,"n_ran_checked":6,"n_instrument":4,"n_unverified":3,"n_honours":0,"n_violates":0,"n_no_contract":6,"n_pointer_only":3,"phrase":"10 ran (of which 0 constructed an object rather than computing a result; 6 with no instrument failure: 0 honoured, 0 violated, 6 with no contract checked; 4 where Syntology's instrument failed) · 3 unverified","sample_list":"/paper/can-editing-llms-inject-harm#ran","syntology_url":"https://syntology.ai/paper/2407.20224","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2407.20224"}},"official":null}},{"url":"/paper/course-correction-safety-alignment-using","slug":"course-correction-safety-alignment-using","title":"Course-Correction: Safety Alignment Using Synthetic Preferences","date":"2024-07-23","arxiv_id":"2407.16637","repositories_listed":1,"syntology":{"n":1,"n_ran":0,"n_constructed":0,"n_ran_checked":0,"n_instrument":0,"n_unverified":1,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":1,"phrase":"0 ran · 1 unverified","sample_list":"/paper/course-correction-safety-alignment-using#ran","syntology_url":"https://syntology.ai/paper/2407.16637","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2407.16637"}},"official":{"repos":["pillowsofwind/course-correction"],"state":"official: harvested, nothing ran","n_ran":0,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":1,"ran_from_kinds":[]}}},{"url":"/paper/the-better-angels-of-machine-personality-how","slug":"the-better-angels-of-machine-personality-how","title":"The Better Angels of Machine Personality: How Personality Relates to LLM Safety","date":"2024-07-17","arxiv_id":"2407.12344","repositories_listed":1,"syntology":null},{"url":"/paper/q-adapter-training-your-llm-adapter-as-a","slug":"q-adapter-training-your-llm-adapter-as-a","title":"Q-Adapter: Customizing Pre-trained LLMs to New Preferences with Forgetting Mitigation","date":"2024-07-04","arxiv_id":"2407.03856","repositories_listed":1,"syntology":{"n":14,"n_ran":13,"n_constructed":0,"n_ran_checked":6,"n_instrument":7,"n_unverified":1,"n_honours":0,"n_violates":0,"n_no_contract":6,"n_pointer_only":1,"phrase":"13 ran (of which 0 constructed an object rather than computing a result; 6 with no instrument failure: 0 honoured, 0 violated, 6 with no contract checked; 7 where Syntology's instrument failed) · 1 unverified","sample_list":"/paper/q-adapter-training-your-llm-adapter-as-a#ran","syntology_url":"https://syntology.ai/paper/2407.03856","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2407.03856"}},"official":{"repos":["mansicer/Q-Adapter"],"state":"official (archive's flag): 2 ran","n_ran":2,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":0,"ran_from_kinds":["found_in_text","official"]}}},{"url":"/paper/safe-unlearning-a-surprisingly-effective-and","slug":"safe-unlearning-a-surprisingly-effective-and","title":"From Theft to Bomb-Making: The Ripple Effect of Unlearning in Defending Against Jailbreak Attacks","date":"2024-07-03","arxiv_id":"2407.02855","repositories_listed":1,"syntology":{"n":7,"n_ran":1,"n_constructed":0,"n_ran_checked":1,"n_instrument":0,"n_unverified":6,"n_honours":0,"n_violates":0,"n_no_contract":1,"n_pointer_only":7,"phrase":"1 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 0 where Syntology's instrument failed) · 6 unverified","sample_list":"/paper/safe-unlearning-a-surprisingly-effective-and#ran","syntology_url":"https://syntology.ai/paper/2407.02855","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2407.02855"}},"official":{"repos":["thu-coai/safeunlearning"],"state":"official (archive's flag): 1 ran","n_ran":1,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":6,"ran_from_kinds":["official"]}}},{"url":"/paper/sop-unlock-the-power-of-social-facilitation","slug":"sop-unlock-the-power-of-social-facilitation","title":"SeqAR: Jailbreak LLMs with Sequential Auto-Generated Characters","date":"2024-07-02","arxiv_id":"2407.01902","repositories_listed":1,"syntology":{"n":1,"n_ran":1,"n_constructed":0,"n_ran_checked":0,"n_instrument":1,"n_unverified":0,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":1,"phrase":"1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/sop-unlock-the-power-of-social-facilitation#ran","syntology_url":"https://syntology.ai/paper/2407.01902","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2407.01902"}},"official":{"repos":["yang-yan-yang-yan/sop"],"state":"official (archive's flag): 1 ran","n_ran":1,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":0,"ran_from_kinds":["official"]}}},{"url":"/paper/safealigner-safety-alignment-against","slug":"safealigner-safety-alignment-against","title":"SafeAligner: Safety Alignment against Jailbreak Attacks via Response Disparity Guidance","date":"2024-06-26","arxiv_id":"2406.18118","repositories_listed":1,"syntology":null},{"url":"/paper/cross-modality-safety-alignment","slug":"cross-modality-safety-alignment","title":"Cross-Modality Safety Alignment","date":"2024-06-21","arxiv_id":"2406.15279","repositories_listed":1,"syntology":{"n":4,"n_ran":4,"n_constructed":0,"n_ran_checked":1,"n_instrument":3,"n_unverified":0,"n_honours":0,"n_violates":1,"n_no_contract":0,"n_pointer_only":4,"phrase":"4 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 1 violated, 0 with no contract checked; 3 where Syntology's instrument failed) · 0 unverified","sample_list":"/paper/cross-modality-safety-alignment#ran","syntology_url":"https://syntology.ai/paper/2406.15279","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2406.15279"}},"official":{"repos":["sinwang20/siuo"],"state":"official (archive's flag): 3 ran","n_ran":3,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":0,"ran_from_kinds":["official","unlocated"]}}},{"url":"/paper/safesora-towards-safety-alignment-of","slug":"safesora-towards-safety-alignment-of","title":"SafeSora: Towards Safety Alignment of Text2Video Generation via a Human Preference Dataset","date":"2024-06-20","arxiv_id":"2406.14477","repositories_listed":1,"syntology":{"n":2,"n_ran":0,"n_constructed":0,"n_ran_checked":0,"n_instrument":0,"n_unverified":2,"n_honours":0,"n_violates":0,"n_no_contract":0,"n_pointer_only":2,"phrase":"0 ran · 2 unverified","sample_list":"/paper/safesora-towards-safety-alignment-of#ran","syntology_url":"https://syntology.ai/paper/2406.14477","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2406.14477"}},"official":{"repos":["pku-alignment/safe-sora"],"state":"official: harvested, nothing ran","n_ran":0,"n_constructed":0,"n_ran_no_instrument_failure":0,"n_unverified":2,"ran_from_kinds":[]}}},{"url":"/paper/safeinfer-context-adaptive-decoding-time","slug":"safeinfer-context-adaptive-decoding-time","title":"SafeInfer: Context Adaptive Decoding Time Safety Alignment for Large Language Models","date":"2024-06-18","arxiv_id":"2406.12274","repositories_listed":1,"syntology":null}],"syntology_records":45,"syntology_note":"a paper without a record is not a recorded non-run: it may lack an arXiv id or simply be absent from the graph layer"},"key_notes":{"n_ran_checked":"legacy name, kept unchanged so existing readers do not break: it counts the samples that ran with no instrument failure (honoured, violated, and ran with no contract checked); it does not mean a contract was checked, and the pages print it as 'K with no instrument failure', not 'K checked'","n_constructed":"a sub-count of the samples that ran, never subtracted from them and never a failure: an executed sample whose run returned an instance of its own class (fixture_out_type equals the entry name): the run built an object and did not compute a result (Syntology's RAN record, counts.constructed)"},"since_archive_judged":{"papers":0,"papers_placed":0,"judged_by":{"model":"a model","models_as_recorded":[],"from":"lb_rows_run_verified.jsonl: each record's `model` field (the extractor's call that wrote decisions[])"},"note":"a model's judgement from the paper's own tables that they report on this leaderboard, not a result: no value from a paper that was not placed is printed or carried here; papers counts each paper once over these leaderboards","backfill":"8,886 of 18,637 newer papers checked so far (733 more had nothing to place; 41 could not be read)","backfill_complete":false,"leaderboards":[]},"papers_where_code_ran":73,"papers_where_code_ran_first_page":"/task/safety-alignment/papers/ran/1","papers_where_code_ran_caption":"We ran code from the paper's repository; we did not run it on this task or check it against the task's benchmarks.","pages":[{"order":"archive","page":2,"of":3,"url":"/task/safety-alignment/papers/2","json":"/task/safety-alignment/papers/2.json","rows":100},{"order":"archive","page":3,"of":3,"url":"/task/safety-alignment/papers/3","json":"/task/safety-alignment/papers/3.json","rows":88},{"order":"ran","page":1,"of":1,"url":"/task/safety-alignment/papers/ran/1","json":"/task/safety-alignment/papers/ran/1.json","rows":73}],"description_links":{"kept":0,"unwrapped_to_text":0,"bare_urls_linked":0,"relative_images_dropped":0,"rule":"internal links are kept only when the target slug exists in the catalog"},"syntology":{"read_at":"2026-10-05T00:36:11+00:00","claim":"Per-sample execution status on synthesized fixtures ('ran N of M samples'); not a correctness claim and not a ranking signal.","status_vocabulary":{"ran_honours":"ran, honoured the contract we drafted","ran_violates":"ran, violated the contract we drafted","ran_draft_wrong":"ran; our contract draft was wrong, not the code","ran_fixture":"ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)","ran":"ran on a synthesized input","unverified":"unverified (harvested, no recorded run)"},"row_route":"each syntology object on a benchmark's first row and on a paper carries sample_list (this site's paper page, #ran), syntology_url (syntology.ai's page for the paper, only when that route probed live) and mcp (the agent call that lists the paper's harvested samples; not a verdict on the paper)","developers":"https://syntology.ai/developers"},"not_shown":{"libraries":"the archive has no per-task library table","trend_sparklines":"the Trend column of the benchmarks table was a rendered image; it is not in the archive","social_and_latest_sorts":"stars and social signals are not in the archive"},"record_sha256":"00364777dca9ce5044fe71a835389985f3b394771e059b1358913280c7e6330a","record_changed_at":"2026-10-05","record_changed_at_basis":"changed"}