Papers › Finding Safety Neurons in Large Language Models

Finding Safety Neurons in Large Language Models

20 Jun 2024arXiv:2406.14144archive 2025-07-28

Jianhui Chen, Xiaozhi Wang, Zijun Yao, Yushi Bai, Lei Hou, Juanzi Li

Large language models (LLMs) excel in various capabilities but also pose safety risks such as generating harmful content and misinformation, even after safety alignment. In this paper, we explore the inner mechanisms of safety alignment from the perspective of mechanistic interpretability, focusing on identifying and analyzing safety neurons within LLMs that are responsible for safety behaviors. We propose generation-time activation contrasting to locate these neurons and dynamic activation patching to evaluate their causal effects. Experiments on multiple recent LLMs show that: (1) Safety neurons are sparse and effective. We can restore 90% safety performance with intervention only on about 5% of all the neurons. (2) Safety neurons encode transferrable mechanisms. They exhibit consistent effectiveness on different red-teaming datasets. The finding of safety neurons also interprets "alignment tax". We observe that the identified key neurons for safety and helpfulness significantly overlap, but they require different activation patterns of the shared neurons. Furthermore, we demonstrate an application of safety neurons in detecting unsafe outputs before generation. Our findings may promote further research on understanding LLM alignment. The source codes will be publicly released to facilitate future research.

PaperPDFCode Syntology ran

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 14 code samples harvested from 1 repository linked to this paper: 14 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; 8 where Syntology's instrument failed) · 0 unverified. Where Syntology's instrument failed: 5 where our contract draft was wrong; 3 where the contract check could not run; failures of Syntology's instrument, not of the code.

By repository: found in paper text by Syntology: 14 samples from 1 repository, 14 ran (6 with no instrument failure; 8 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

Levels written between 2026-09-24 and 2026-09-26 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed. 3 of the 14 samples that ran have no recorded date.

THU-KEG/SafetyNeuron found in paper text by SyntologyMITSyntology: 14 samples, 14 ran (8 instrument-failed) · 0 unverified report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

14 samples harvested: 14 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; 8 where Syntology's instrument failed) · 0 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

5ran · our draft was wrong
3ran; the contract check could not run
6ran

Licence: 1 of the 14 samples is pointer only, meaning Syntology does not serve that copy's text. For that sample, this paper's copy is under a permissive licence and cleared, but the stored copy of the same code that Syntology serves is not; the licence label's tooltip gives Syntology's recorded reason. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from THU-KEG/SafetyNeuron. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity. A census tag appears when Syntology's census ran implementations with that name on one shared input per bucket: it gives how many were compared, from how many papers, and how many distinct outputs they produced, and names this sample's bucket when this sample was one of them.

THU-KEG/SafetyNeuron found in paper text by Syntology 14 samples, 14 ran (8 instrument-failed) · 0 unverified
repeat_kv in 244 papers
census: 15 implementations from 60 papers, 2 distinct outputs on one shared input this sample: bucket 1 of 1, arg 1: rank 4, kind float, dtype float32
src/models/HookedGemma.py ran; the contract check could not run (RAISES) · metamorphic tier: invariant fingerprinted MIT (permissive) · 30d7eec482ebf6b1 · report fixture: signature hidden_states:Tensor/4/float, n_rep:int/0/int · returned Tensor · level recorded, date not recorded
apply_rotary_pos_emb in 393 papers
census: 14 implementations from 15 papers, 8 distinct outputs across 7 buckets, one shared input per bucket
src/models/HookedGemma.py ran · our draft was wrong MIT (permissive) · bac65c3dafaec040 · report fixture: signature q:Tensor/4/float, k:Tensor/4/float, cos:Tensor/3/float, sin:Tensor/3/float, position_ids:NoneType/0/?, unsqueeze_dim:int/0/int · returned tuple · level recorded, date not recorded
create_prompt_with_llama2_chat_format in 11 papers src/eval/templates.py ran · our draft was wrong MIT (permissive) · e51c9713fadad16f · report fixture: signature messages:list/1/dict, bos:str/0/str, eos:str/0/str, add_bos:bool/0/bool · returned str · level recorded 2026-09-26
create_prompt_with_tulu_chat_format in 13 papers src/eval/templates.py ran MIT (permissive) here; the stored copy is pointer only · 697d92a5d51eff3e · report fixture: signature messages:list/1/dict, bos:str/0/str, eos:str/0/str, add_bos:bool/0/bool · returned str · level recorded 2026-09-25
dispatch_openai_chat_requesets in 12 papers src/eval/dispatch_openai_requests.py ran MIT (permissive) · 233355240185dc1a · report fixture: signature messages_list:list/1/list, model:str/0/str, temperature:float/0/float, max_tokens:int/0/int, top_p:float/0/float · returned coroutine · level recorded 2026-09-24
dispatch_openai_prompt_requesets in 12 papers src/eval/dispatch_openai_requests.py ran · our draft was wrong MIT (permissive) · 0979c8368e7c6a08 · report fixture: signature prompt_list:list/1/str, model:str/0/str, completion_kwargs:dict/0/? · returned coroutine · level recorded 2026-09-24
generate_completions in 38 papers src/eval/utils.py ran; the contract check could not run (RAISES) MIT (permissive) · 4b7fca0051f25274 · report fixture: signature model:callable/0/?, tokenizer:callable/0/?, prompts:list/1/str, batch_size:int/0/int, stop_id_sequences:list/1/list, add_special_tokens:bool/0/bool, disable_tqdm:bool/0/bool, num_return_sequences:int/0/int, max_new_tokens:int/0/int · returned list · level recorded 2026-09-24
generate_completions_and_masks src/eval/utils.py ran MIT (permissive) · b4722271cba47572 · report fixture: signature model:callable/0/?, tokenizer:callable/0/?, prompts:list/1/str, batch_size:int/0/int, add_special_tokens:bool/0/bool, disable_tqdm:bool/0/bool, num_return_sequences:int/0/int, max_length:int/0/int · returned tuple · level recorded 2026-09-24
get_act_name src/utils.py ran fingerprinted MIT (permissive) · 0d845b84c5219607 · report fixture: signature name:str/0/str, layer:int/0/int, layer_type:str/0/str · returned str · level recorded 2026-09-24
get_clf src/predict_before_gen.py ran MIT (permissive) · 0f88e41d70d1279d · report fixture: signature peft:str/0/str, dataset:str/0/str, topk:int/0/int · returned LogisticRegression · level recorded 2026-09-24
layer_patch_hook src/ppl.py ran MIT (permissive) · 508cbcd9d0571483 · report fixture: signature value:Tensor/3/float, hook:str/0/str, neurons:list/1/int, patched_values:Tensor/3/float · returned Tensor · level recorded 2026-09-24
load_adapter_weights src/utils.py ran · our draft was wrong MIT (permissive) · 34cedc62fcf7195b · report fixture: signature path:str/0/str · returned dict · level recorded 2026-09-24
rotate_half in 424 papers
census: 24 implementations from 96 papers, 4 distinct outputs across 2 buckets, one shared input per bucket this sample: bucket 1 of 2, arg 1: rank 2, kind float, dtype float32
src/models/HookedGemma.py ran · our draft was wrong fingerprinted MIT (permissive) · b99eea6376d1e212 · report fixture: signature x:Tensor/2/float · returned Tensor · level recorded, date not recorded
topk_index src/utils.py ran; the contract check could not run (RAISES) MIT (permissive) · 8ae41f0bacbff51f · report fixture: signature tensor:Tensor/2/float, k:int/0/int · returned zip · level recorded 2026-09-24

Tasks

MisinformationRed TeamingSafety Alignment

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

Patching

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections