Skip to content
OpenTrain AIFor AI Companies

HFEPX · Eval paper review

Self-Preference Is Weak or Absent in Verifiable Instruction-Following Revision: A Four-Model Test Under Genuine Authorship

William Guey, Pierrick Bougault

Published

Jun 18, 2026

Citations

0

Trust level

Moderate

Usefulness score

50/100 (Medium)

Extraction confidence

55% (Moderate)

Derived from extracted protocol signals and abstract evidence.

Rater population

Not reported

Signals refreshed

Jun 18, 2026

Open arXiv Find implementation Shortlist (0) RSS feed

Tags

Critique Edit Law Pairwise Preference

Should you rely on this paper?

This paper has useful evaluation signal, but protocol completeness is partial; pair it with related papers before deciding implementation strategy.

Use this for comparison and orientation, not as your only source.

Best use

Secondary protocol comparison source

Use if you need

Background context only.

What to verify

Validate the evaluation procedure and quality controls in the full paper before operational use.

Main weakness

The abstract does not clearly describe the evaluation setup.

Human feedback signal
Detected
From extracted signals
Evaluation signal
Weak or implicit
Validate from full paper
Usefulness for eval research
50/100
Moderate-confidence candidate

Useful as a secondary reference; validate protocol details against neighboring papers.

If you are doing eval pipeline work, start here

Human Eval Papers LLM-as-Judge Papers Pairwise Preference Papers Tool-Use Evaluation Search

Abstract

Large language models (LLMs) increasingly review and revise text, including their own. A documented self-preference bias (models favoring their own generations when acting as judges) raises the question of whether models also resist valid corrections to their own writing. We test this in a setting where "valid" is decided not by another model but by a deterministic verifier: instruction-following revision on IFEval. A model writes a draft; the official IFEval checker confirms the draft violates a constraint and that a candidate edit fixes it; the model then accepts or rejects that edit either as the genuine in-context author or as a fresh model that sees the draft neutrally. Across four mid-tier model families and 85 author-versus-fresh comparisons, we find no detectable self-preference: authors reject verified-good fixes to their own drafts at essentially the same rate as fresh models judging the same drafts (gap -5.1 pp, 95% CI [-12.9, +2.7]). A self-skepticism hint from a smaller pilot did not replicate at scale. The one robust observation is qualitative: when authors do reject a verified-good fix, 97% of their stated reasons are flaw-catching rather than preference, that is, about the character of rejections, not an elevated rate. Effects smaller than ~13 pp cannot be excluded at this sample size.

What we could verify

These are the protocol signals we could actually recover from the available paper metadata. Use them to decide whether this paper is worth deeper reading.

Human Feedback Types

strong

Pairwise Preference, Critique Edit

Directly usable for protocol triage.

"Large language models (LLMs) increasingly review and revise text, including their own."

Evaluation Modes

missing

None explicit

Validate eval design from full paper text.

"Large language models (LLMs) increasingly review and revise text, including their own."

Quality Controls

missing

Not reported

No explicit QC controls found.

"Large language models (LLMs) increasingly review and revise text, including their own."

Benchmarks / Datasets

strong

IFEval

Useful for quick benchmark comparison.

"We test this in a setting where "valid" is decided not by another model but by a deterministic verifier: instruction-following revision on IFEval."

Reported Metrics

missing

Not extracted

No metric anchors detected.

"Large language models (LLMs) increasingly review and revise text, including their own."

Benchmarks and datasets

IFEval

Reported metrics

No metric terms were extracted from the available abstract.

Human feedback details
Uses human feedback
Yes
Feedback types
Pairwise Preference, Critique Edit
Rater population
Not reported
Expertise required
General
Evaluation details
Evaluation modes
None
Agentic eval
None
Quality controls
Not reported
Evidence quality
Moderate
Use this page as
Secondary protocol comparison source

Research brief

Metadata summary

Large language models (LLMs) increasingly review and revise text, including their own.

Based on abstract + metadata only. Check the source paper before making high-confidence protocol decisions.

Key takeaways

  • Large language models (LLMs) increasingly review and revise text, including their own.
  • A documented self-preference bias (models favoring their own generations when acting as judges) raises the question of whether models also resist valid corrections to their own writing.
  • We test this in a setting where "valid" is decided not by another model but by a deterministic verifier: instruction-following revision on IFEval.

Researcher actions

  • Compare this paper against nearby papers in the same arXiv category before using it for protocol decisions.
  • Check the full text for explicit evaluation design choices (raters, protocol, and metrics).
  • Use related-paper links to find stronger protocol-specific references.

Caveats

  • Generated from abstract + metadata only; no PDF parsing.
  • Signals below are heuristic and may miss details reported outside the abstract.

Recommended queries

Pairwise preference evaluation

Contribution summary

  • A documented self-preference bias (models favoring their own generations when acting as judges) raises the question of whether models also resist valid corrections to their own writing.
  • Across four mid-tier model families and 85 author-versus-fresh comparisons, we find no detectable self-preference: authors reject verified-good fixes to their own drafts at essentially the same rate as fresh models judging the same drafts…
  • The one robust observation is qualitative: when authors do reject a verified-good fix, 97% of their stated reasons are flaw-catching rather than preference, that is, about the character of rejections, not an elevated rate.

Why it matters for eval

  • Across four mid-tier model families and 85 author-versus-fresh comparisons, we find no detectable self-preference: authors reject verified-good fixes to their own drafts at essentially the same rate as fresh models judging the same drafts…
  • The one robust observation is qualitative: when authors do reject a verified-good fix, 97% of their stated reasons are flaw-catching rather than preference, that is, about the character of rejections, not an elevated rate.

Researcher checklist

  • Human feedback protocol is explicit

    Detected: Pairwise Preference, Critique Edit

  • Evaluation mode is explicit

    No clear evaluation mode extracted.

  • Quality control reporting appears

    No calibration/adjudication/IAA control explicitly detected.

  • Benchmark or dataset anchors are present

    Detected: IFEval

  • Metric reporting is present

    No metric terms extracted.