Papers › Fairer Preferences Elicit Improved Human-Aligned Large Language Model Judgments

Fairer Preferences Elicit Improved Human-Aligned Large Language Model Judgments

17 Jun 2024arXiv:2406.11370archive 2025-07-28

Han Zhou, Xingchen Wan, Yinhong Liu, Nigel Collier, Ivan Vulić, Anna Korhonen

Large language models (LLMs) have shown promising abilities as cost-effective and reference-free evaluators for assessing language generation quality. In particular, pairwise LLM evaluators, which compare two generated texts and determine the preferred one, have been employed in a wide range of applications. However, LLMs exhibit preference biases and worrying sensitivity to prompt designs. In this work, we first reveal that the predictive preference of LLMs can be highly brittle and skewed, even with semantically equivalent instructions. We find that fairer predictive preferences from LLMs consistently lead to judgments that are better aligned with humans. Motivated by this phenomenon, we propose an automatic Zero-shot Evaluation-oriented Prompt Optimization framework, ZEPO, which aims to produce fairer preference decisions and improve the alignment of LLM evaluators with human judgments. To this end, we propose a zero-shot learning objective based on the preference decision fairness. ZEPO demonstrates substantial performance improvements over state-of-the-art LLM evaluators, without requiring labeled data, on representative meta-evaluation benchmarks. Our findings underscore the critical correlation between preference fairness and human alignment, positioning ZEPO as an efficient prompt optimizer for bridging the gap between LLM evaluators and human judgments.

PaperPDFCodeCode Syntology ranMethod as code

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 29 code samples harvested from 2 repositories linked to this paper: 20 ran (of which 4 constructed an object rather than computing a result; 10 with no instrument failure: 3 honoured, 2 violated, 5 with no contract checked; 10 where Syntology's instrument failed) · 9 unverified. Where Syntology's instrument failed: 10 where our contract draft was wrong; failures of Syntology's instrument, not of the code.

By repository: official repository: 23 samples from 1 repository, 17 ran (7 with no instrument failure; 10 where Syntology's instrument failed); community (archive-listed): 6 samples from 1 repository, 3 ran (3 with no instrument failure; 0 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 on 2026-09-22 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

cambridgeltl/zepo officialmentioned in papermentioned on GitHubpytorchSyntology: 23 samples, 17 ran (10 instrument-failed) · 6 unverified report
cambridgeltl/pairs mentioned on GitHubpytorchSyntology: 6 samples, 3 ran (0 instrument-failed) · 3 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

29 samples harvested: 20 ran (of which 4 constructed an object rather than computing a result; 10 with no instrument failure: 3 honoured, 2 violated, 5 with no contract checked; 10 where Syntology's instrument failed) · 9 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

3ran · honoured contract
2ran · violated contract
10ran · our draft was wrong
5ran
9unverified

Licence: 2 of the 29 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from 2 repositories linked to this paper, official or community; the list below is grouped by repository, each group headed by its name and what it is to the paper. 2 samples whose link to this paper records no repository are listed under one of this paper's own repositories, because identical code was harvested from it for another paper. “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.

cambridgeltl/zepo official repository 23 samples, 17 ran (10 instrument-failed) · 6 unverified
CompareResultObject zepo.py ran MIT (permissive) · fa3627b39b2fc64c · report fixture: returned CompareResultObject · level recorded 2026-09-22
OpenAIChatModel zepo.py ran MIT (permissive) · e08a10d84561235c · report fixture: returned OpenAIChatModel · level recorded 2026-09-22
Timer in 12 papers zepo.py ran MIT (permissive) · 39083ac397faebe6 · report fixture: returned Timer · level recorded 2026-09-22
calculate_uncertainty in 37 papers zepo.py ran · honoured contract fingerprinted MIT (permissive) · ca99032ac77b29e6 · report fixture: signature probablities:list/1/float · returned float64 · level recorded 2026-09-22
compute_pairwise_preference_matrix zepo.py ran · our draft was wrong MIT (permissive) · 6788ac0f668741a9 · report fixture: signature model:MockComparisonModel/0/?, input:str/0/str, output:list/1/str, prompt_templates:list/1/Template, do_cot:bool/0/bool, worker_num:int/0/int, batch_size:int/0/int, instruction:NoneType/0/? · returned tuple · level recorded 2026-09-22
get_cot_compare_prompt_template zepo.py ran · our draft was wrong MIT (permissive) · e832e11e0d101a37 · report fixture: signature dataset:str/0/str · returned str · level recorded 2026-09-22
get_cot_eval_prompt_template zepo.py ran · metamorphic tier: deterministic MIT (permissive) · 72e4f7f276e2410b · report fixture: returned str · level recorded 2026-09-22
get_pairwise_prompt_template zepo.py ran · our draft was wrong MIT (permissive) · 4ed979529f4db748 · report fixture: signature dataset:str/0/str, use_instruction:bool/0/bool · returned str · level recorded 2026-09-22
is_integer_string zepo.py ran · violated contract MIT (permissive) · 4419dd4c768a4bb5 · report fixture: signature s:str/0/str · returned bool · level recorded 2026-09-22
list_to_pairwise_non_diagonal zepo.py ran · our draft was wrong fingerprinted MIT (permissive) · 4a77046e8d79d3e6 · report fixture: signature non_diagonal_list:list/1/int, size:int/0/int · returned ndarray · level recorded 2026-09-22
load_TopicalChat zepo.py ran · our draft was wrong MIT (permissive) · c3b4a5412489fb24 · report fixture: signature path:str/0/str, truncate_num_for_eval:NoneType/0/? · returned tuple · level recorded 2026-09-22
load_gsm8k zepo.py ran · our draft was wrong MIT (permissive) · 99dff92abeb97011 · report fixture: signature data_path:str/0/str, cot:bool/0/bool · returned tuple · level recorded 2026-09-22
load_json in 344 papers zepo.py ran · our draft was wrong MIT (permissive) · 4bf2c02de1d92773 · report fixture: signature file_path:str/0/str · returned dict · level recorded 2026-09-22
load_jsonl in 179 papers zepo.py ran · our draft was wrong MIT (permissive) · 472d563b979e2b49 · report fixture: signature file_path:str/0/str · returned list · level recorded 2026-09-22
load_newsroom zepo.py ran · our draft was wrong MIT (permissive) · 1841f2ca10ba4b50 · report fixture: signature path:str/0/str, flat_output:bool/0/bool, truncate_num_for_eval:NoneType/0/? · returned tuple · level recorded 2026-09-22
load_summEval zepo.py ran · our draft was wrong MIT (permissive) · b5816ff43e6dc12a · report fixture: signature path:str/0/str, flat_output:bool/0/bool, truncate_num_for_eval:NoneType/0/? · returned tuple · level recorded 2026-09-22
pairwise_non_diagonal_to_list zepo.py ran · honoured contract MIT (permissive) · 5c7d207f402ec83c · report fixture: signature size:int/0/int · returned list · level recorded 2026-09-22
Llama2ModelLocal zepo.py unverified MIT (permissive) · 3b641c847f128e7f · report
Llama3ModelLocal zepo.py unverified MIT (permissive) · b9664c08870bc614 · report
MistralModelLocal zepo.py unverified MIT (permissive) · 7b238544c212100b · report
get_instruction in 10 papers zepo.py unverified MIT (permissive) · 902f7af3019415e0 · report
pairwise_compare zepo.py unverified MIT (permissive) · 9a7434c691505bde · report
zepo zepo.py unverified MIT (permissive) · 421bae01e07ebb79 · report
cambridgeltl/pairs community (archive-listed) 6 samples, 3 ran (0 instrument-failed) · 3 unverified
BeamItem pairs/pairs_ranking.py ran MIT (permissive) · de96e87fcdc09240 · report fixture: returned BeamItem · level recorded 2026-09-22
get_likelihood_coefficient pairs/pairs_ranking.py identical code Syntology harvested from this repository for another paper; this paper's link records no repository, path or licence of its own ran · honoured contract licence of this copy not recorded · 14e3151e6acdfedb · report fixture: signature N:int/0/int, p:float/0/float · returned float64 · level recorded 2026-09-22
moving_average in 46 papers
census: 9 implementations from 11 papers, 6 distinct outputs across 4 buckets, one shared input per bucket this sample: bucket 2 of 4, arg 1: rank 0, kind float, dtype float32 · arg 2: rank 0, kind float, dtype float32
pairs/pairs_ranking.py identical code Syntology harvested from this repository for another paper; this paper's link records no repository, path or licence of its own ran · violated contract fingerprinted licence of this copy not recorded · 9edbea252c4a2b7e · report fixture: signature sum:Tensor/0/float, val:Tensor/0/float, idx:int/0/int · returned Tensor · level recorded 2026-09-22
PairsBeam pairs/pairs_ranking.py unverified MIT (permissive) · a0af76fa6f8f0944 · report
is_better_than_prob pairs/pairs_ranking.py unverified MIT (permissive) · 63ce88ca435a0f62 · report
merge_with_confidence_beam pairs/pairs_ranking.py unverified MIT (permissive) · 92280df256c9d8b0 · report

Method as code Syntology

What Syntology's graph records as this paper's claimed method, lifted into code: each assembly names its bricks (functions and classes), each with its own role and its own verification level. An assembly is not a check that the code implements the method correctly or reproduces the paper's results. Levels are each brick's own verification level in Syntology's graph (level 0: no recorded run), listed brick by brick and never combined into one level; “ran” means the brick executed on a synthesized input, not that its output is correct, and the instrument figure counts failures of Syntology's instrument, not of the code. Where a model read the code against the paper's claim, each assembly's label verdict is one model's judgement, kept in this page's JSON twin; it is not calibrated, so no card prints it. Siblings are assemblies lifted from this paper in the same batch for the same method: alternative readings kept side by side; none replaces or supersedes another.

2 assemblies on this paper: 2 harvested, 0 generated, listed in 1 groups (siblings lifted in one batch form one group).

2 sibling assemblies for ZEPO (harvested), lifted in the same batch on 2026-09-09: alternative readings kept side by side; none replaces or supersedes another.

harvested Assembly for ZEPO (graph method: ZEPO). 23 bricks; individually on a fixture: 17 ran (of which 3 constructed an object rather than computing a result; 7 with no instrument failure: 2 honoured, 1 violated, 4 with no contract checked; 10 where Syntology's instrument failed) · 6 unverified. The assembly has not been run end-to-end.

one of 2 siblings

  1. zepo entry · level 0 · unverified · 421bae01e07ebb79
  2. get_instruction helper · level 0 · unverified · 902f7af3019415e0
  3. pairwise_compare helper · level 0 · unverified · 9a7434c691505bde
  4. OpenAIChatModel helper · level 1 · ran · e08a10d84561235c
  5. Llama2ModelLocal helper · level 0 · unverified · 3b641c847f128e7f
  6. Llama3ModelLocal helper · level 0 · unverified · b9664c08870bc614
  7. MistralModelLocal helper · level 0 · unverified · 7b238544c212100b
  8. compute_pairwise_preference_matrix metric · level 1 · ran · our draft was wrong · 6788ac0f668741a9
  9. get_cot_compare_prompt_template helper · level 1 · ran · our draft was wrong · e832e11e0d101a37
  10. get_cot_eval_prompt_template helper · level 1 · ran · 72e4f7f276e2410b
  11. get_pairwise_prompt_template helper · level 1 · ran · our draft was wrong · 4ed979529f4db748
  12. load_TopicalChat helper · level 1 · ran · our draft was wrong · c3b4a5412489fb24
  13. load_gsm8k helper · level 1 · ran · our draft was wrong · 99dff92abeb97011
  14. load_newsroom helper · level 1 · ran · our draft was wrong · 1841f2ca10ba4b50
  15. load_summEval helper · level 1 · ran · our draft was wrong · b5816ff43e6dc12a
  16. Timer helper · level 1 · ran · 39083ac397faebe6
  17. CompareResultObject helper · level 1 · ran · fa3627b39b2fc64c
  18. calculate_uncertainty helper · level 1 · ran · honoured contract · ca99032ac77b29e6
  19. is_integer_string helper · level 1 · ran · violated contract · 4419dd4c768a4bb5
  20. list_to_pairwise_non_diagonal helper · level 1 · ran · our draft was wrong · 4a77046e8d79d3e6
  21. pairwise_non_diagonal_to_list helper · level 1 · ran · honoured contract · 5c7d207f402ec83c
  22. load_json helper · level 1 · ran · our draft was wrong · 4bf2c02de1d92773
  23. load_jsonl helper · level 1 · ran · our draft was wrong · 472d563b979e2b49

harvested Assembly for ZEPO (graph method: ZEPO). 6 bricks; individually on a fixture: 3 ran (of which 1 constructed an object rather than computing a result; 3 with no instrument failure: 1 honoured, 1 violated, 1 with no contract checked; 0 where Syntology's instrument failed) · 3 unverified. The assembly has not been run end-to-end.

one of 2 siblings

  1. PairsBeam entry · level 0 · unverified · a0af76fa6f8f0944
  2. merge_with_confidence_beam helper · level 0 · unverified · 92280df256c9d8b0
  3. BeamItem helper · level 1 · ran · de96e87fcdc09240
  4. get_likelihood_coefficient helper · level 1 · ran · honoured contract · 14e3151e6acdfedb
  5. is_better_than_prob helper · level 0 · unverified · 63ce88ca435a0f62
  6. moving_average helper · level 1 · ran · violated contract · 9edbea252c4a2b7e

Each brick ends with its code_sha256, Syntology's identity for that exact code; click one to copy the get_code call that returns it. For agents, on Syntology's MCP service: assembly(arxiv_id="2406.11370") returns these assemblies with each brick's role and wiring (how to connect).

Tasks

FairnessLanguage ModelingLanguage ModellingLarge Language ModelText GenerationZero-Shot Learning

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

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