Skip to content
OpenTrain AIFor AI Companies

HFEPX · Eval paper review

Is the Judge Its Twin? Post-training makes a model's writing more predictable but barely moves its taste, as a judge, toward predictable writing

Arman Nik Khah, Arvin Bahreini

Published

Sep 26, 2026

Citations

0

Trust level

Moderate

Usefulness score

57/100 (Medium)

Extraction confidence

65% (Moderate)

Derived from extracted protocol signals and abstract evidence.

Rater population

Not reported

Signals refreshed

Oct 22, 2026

Open arXiv Find implementation Shortlist (0) RSS feed

Tags

General Llm As Judge 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

A secondary eval reference to pair with stronger protocol papers.

What to verify

Read the full paper before copying any benchmark, metric, or protocol choices.

Main weakness

The abstract does not clearly name benchmarks or metrics.

Human feedback signal
Detected
From extracted signals
Evaluation signal
Detected
Eval setup described
Usefulness for eval research
57/100
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

Language models now routinely grade other language models. If post-training makes a model's writing more predictable, it may also teach the same model, as a judge, to reward predictable writing. We test this directly. We follow two open 7-billion-parameter model families, OLMo-2 and Zephyr, from their base models through supervised fine-tuning (SFT) and direct preference optimization (DPO), and measure every stage as a writer of short stories and as a judge of story pairs. As writers, the models drift as feared. Each family's trained model finds its stages' stories more familiar in the order base, SFT, DPO, in all ten prompts, and a model from the other family finds the trained stories 7.0 to 9.2 percent (OLMo-2) and 24 to 25 percent (Zephyr) less surprising per token. As judges, they barely move toward predictable writing. Asked which story is better, a question every judge passes a competence check on, no trained judge's preference for the more predictable story grows by as much as one point of pick probability at its estimate. A one-sided 95% upper bound on that growth is about 5 points on the widest pairs. Training instead makes judges more swayed by story length and by display order, and it breaks "more creative" as a question. Trained judges no longer reliably prefer a story to its own words in scrambled order. For these two families, the practical hazard in a small trained judge of creative writing lies in these measurable habits.

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

Directly usable for protocol triage.

"Language models now routinely grade other language models."

Evaluation Modes

strong

Llm As Judge

Includes extracted eval setup.

"Language models now routinely grade other language models."

Quality Controls

missing

Not reported

No explicit QC controls found.

"Language models now routinely grade other language models."

Benchmarks / Datasets

missing

Not extracted

No benchmark anchors detected.

"Language models now routinely grade other language models."

Reported Metrics

missing

Not extracted

No metric anchors detected.

"Language models now routinely grade other language models."

Benchmarks and datasets

No benchmark or dataset names were extracted from the available abstract.

Reported metrics

No metric terms were extracted from the available abstract.

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

Research brief

Metadata summary

Language models now routinely grade other language models.

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

Key takeaways

  • Language models now routinely grade other language models.
  • If post-training makes a model's writing more predictable, it may also teach the same model, as a judge, to reward predictable writing.
  • We follow two open 7-billion-parameter model families, OLMo-2 and Zephyr, from their base models through supervised fine-tuning (SFT) and direct preference optimization (DPO), and measure every stage as a writer of short stories and as a judge of story pairs.

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

  • If post-training makes a model's writing more predictable, it may also teach the same model, as a judge, to reward predictable writing.
  • We follow two open 7-billion-parameter model families, OLMo-2 and Zephyr, from their base models through supervised fine-tuning (SFT) and direct preference optimization (DPO), and measure every stage as a writer of short stories and as a…
  • As judges, they barely move toward predictable writing.

Why it matters for eval

  • If post-training makes a model's writing more predictable, it may also teach the same model, as a judge, to reward predictable writing.
  • We follow two open 7-billion-parameter model families, OLMo-2 and Zephyr, from their base models through supervised fine-tuning (SFT) and direct preference optimization (DPO), and measure every stage as a writer of short stories and as a…

Researcher checklist

  • Human feedback protocol is explicit

    Detected: Pairwise Preference

  • Evaluation mode is explicit

    Detected: Llm As Judge

  • Quality control reporting appears

    No calibration/adjudication/IAA control explicitly detected.

  • Benchmark or dataset anchors are present

    No benchmark/dataset anchor extracted from abstract.

  • Metric reporting is present

    No metric terms extracted.