Pith. sign in

REVIEW 11 cited by

Self-Taught Evaluators

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2408.02666 v2 pith:MU77BDGI submitted 2024-08-05 cs.CL cs.AI

classification cs.CLcs.AI
keywords modeldataevaluatorshumantrainingapproachevaluationimprove
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Model-based evaluation is at the heart of successful model development -- as a reward model for training, and as a replacement for human evaluation. To train such evaluators, the standard approach is to collect a large amount of human preference judgments over model responses, which is costly and the data becomes stale as models improve. In this work, we present an approach that aims to im-prove evaluators without human annotations, using synthetic training data only. Starting from unlabeled instructions, our iterative self-improvement scheme generates contrasting model outputs and trains an LLM-as-a-Judge to produce reasoning traces and final judgments, repeating this training at each new iteration using the improved predictions. Without any labeled preference data, our Self-Taught Evaluator can improve a strong LLM (Llama3-70B-Instruct) from 75.4 to 88.3 (88.7 with majority vote) on RewardBench. This outperforms commonly used LLM judges such as GPT-4 and matches the performance of the top-performing reward models trained with labeled examples.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 11 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. CodeJudgeBench: Benchmarking LLM-as-a-Judge for Coding Tasks

    cs.CL 2025-07 conditional novelty 7.0 of 10

    A 5,352-pair benchmark shows thinking LLMs outperform non-thinking ones as code judges, but position bias and formatting sensitivity remain strong.

  2. Improving Large Vision and Language Models by Learning from a Panel of Peers

    cs.CV 2025-09 conditional novelty 6.0 of 10

    A panel of LVLMs that generate, evaluate, and learn from each other's outputs improves average benchmark scores by 9 points across 15 tasks.

  3. Multilingual Self-Taught Faithfulness Evaluators

    cs.CL 2025-07 conditional novelty 6.0 of 10

    STEMF trains multilingual faithfulness evaluators from synthetic data alone, and English-only training yields the best average results across languages.

  4. Improving Data and Parameter Efficiency of Neural Language Models Using Representation Analysis

    cs.CL 2025-07 conditional novelty 6.0 of 10

    Representation smoothness can be used to regularize training, stop early without validation labels, and guide active learning combined with parameter-efficient fine-tuning, reducing data and compute.

  5. From General to Targeted Rewards: Surpassing GPT-4 in Open-Ended Long-Context Generation

    cs.CL 2025-06 conditional novelty 6.0 of 10

    ProxyReward trains long-form generation models by rewarding how well an AI judge can answer generated yes/no questions about the response, improving open-source models on ProxyQA.

  6. Beyond Gold Standards: Epistemic Ensemble of LLM Judges for Formal Mathematical Reasoning

    cs.CL 2025-06 conditional novelty 6.0 of 10

    A taxonomy-guided ensemble of LLM judges correlates with human ratings of autoformalizations (up to 0.662 on Isabelle/HOL) better than coarse-grained judges and reference metrics, but validation is partly in-sample an...

  7. SkillVerse : Assessing and Enhancing LLMs with Tree Evaluation

    cs.CL 2025-05 conditional novelty 6.0 of 10

    SkillVerse clusters LLM critiques into a dendrogram of skills and uses it to improve in-context learning and predict unseen model weaknesses.

  8. Learning to Reason via Mixture-of-Thought for Logical Reasoning

    cs.CL 2025-05 conditional novelty 6.0 of 10

    Jointly training and voting across natural language, code, and truth-table reasoning modalities improves LLM logical reasoning accuracy by up to 11.7 percentage points.

  9. Codifying the Judge: Scalable Evaluation via Program Distillation

    cs.AI 2026-05 conditional novelty 5.0 of 10

    LLM judge logic can be distilled into a committee of Python scoring programs that match mid-size LLM judge accuracy on filtered preference datasets, at orders-of-magnitude higher throughput.

  10. Advancing Multimodal Judge Models through a Capability-Oriented Benchmark and MCTS-Driven Data Generation

    cs.AI 2026-02 reject novelty 5.0 of 10

    A capability-oriented multimodal judge benchmark and MCTS-based preference-data generation improve judge models on some benchmarks, but the paper's SOTA claim is not supported by its own numbers.

  11. Reward Reasoning Model

    cs.CL 2025-05 conditional novelty 4.0 of 10

    Reward models that think before judging, trained via reinforcement learning without human-written reasoning traces, outperform standard reward models and improve with more test-time compute.

Pith tools