# On scalable oversight with weak LLMs judging strong LLMs

arXiv: https://arxiv.org/abs/2407.04622 · HTML: https://mlsanity.com/paper/2407.04622/

## Key facts

- Cited by 99 papers in the map (data through 14 Oct 2026).
- Builds on 41 of its 71 references that are in the map.
- First posted on arXiv 5 Jul 2024.

## Builds on (most cited first, 30 of 41)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context](https://mlsanity.com/paper/2403.05530/) (2024)
- [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://mlsanity.com/paper/2109.07958/) (2021)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions](https://mlsanity.com/paper/1905.10044/) (2019)
- [MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI](https://mlsanity.com/paper/2311.16502/) (2023)
- [Improving Factuality and Reasoning in Language Models through Multiagent Debate](https://mlsanity.com/paper/2305.14325/) (2023)
- [Concrete Problems in AI Safety](https://mlsanity.com/paper/1606.06565/) (2016)
- [Large Language Models are not Fair Evaluators](https://mlsanity.com/paper/2305.17926/) (2023)
- [Large Language Models Cannot Self-Correct Reasoning Yet](https://mlsanity.com/paper/2310.01798/) (2023)
- [Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate](https://mlsanity.com/paper/2305.19118/) (2023)
- [ChatEval: Towards Better LLM-based Evaluators through Multi-Agent Debate](https://mlsanity.com/paper/2308.07201/) (2023)
- [LLM Evaluators Recognize and Favor Their Own Generations](https://mlsanity.com/paper/2404.13076/) (2024)
- [PAL: Program-aided Language Models](https://mlsanity.com/paper/2211.10435/) (2022)
- [Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision](https://mlsanity.com/paper/2312.09390/) (2023)
- [Scalable agent alignment via reward modeling: a research direction](https://mlsanity.com/paper/1811.07871/) (2018)
- [AI safety via debate](https://mlsanity.com/paper/1805.00899/) (2018)
- [Self-critiquing models for assisting human evaluators](https://mlsanity.com/paper/2206.05802/) (2022)
- [Language Models Are Greedy Reasoners: A Systematic Formal Analysis of Chain-of-Thought](https://mlsanity.com/paper/2210.01240/) (2022)
- [Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Models](https://mlsanity.com/paper/2404.18796/) (2024)
- [Debating with More Persuasive LLMs Leads to More Truthful Answers](https://mlsanity.com/paper/2402.06782/) (2024)
- [Risks from Learned Optimization in Advanced Machine Learning Systems](https://mlsanity.com/paper/1906.01820/) (2019)
- [The Alignment Problem from a Deep Learning Perspective](https://mlsanity.com/paper/2209.00626/) (2022)
- [Measuring Progress on Scalable Oversight for Large Language Models](https://mlsanity.com/paper/2211.03540/) (2022)
- [QuALITY: Question Answering with Long Input Texts, Yes!](https://mlsanity.com/paper/2112.08608/) (2021)
- [Many-Shot In-Context Learning](https://mlsanity.com/paper/2404.11018/) (2024)

## Built on by (most cited first, 30 of 99)

- [From Generation to Judgment: Opportunities and Challenges of LLM-as-a-judge](https://mlsanity.com/paper/2411.16594/) (2024)
- [Talk Isn't Always Cheap: Understanding Failure Modes in Multi-Agent Debate](https://mlsanity.com/paper/2509.05396/) (2025)
- [Hidden in Plain Text: Emergence & Mitigation of Steganographic Collusion in LLMs](https://mlsanity.com/paper/2410.03768/) (2024)
- [Teaching Language Models to Critique via Reinforcement Learning](https://mlsanity.com/paper/2502.03492/) (2025)
- [A Survey on Personalized Alignment -- The Missing Piece for Large Language Models in Real-World Applications](https://mlsanity.com/paper/2503.17003/) (2025)
- [Steering Out-of-Distribution Generalization with Concept Ablation Fine-Tuning](https://mlsanity.com/paper/2507.16795/) (2025)
- [Do LLM Evaluators Prefer Themselves for a Reason?](https://mlsanity.com/paper/2504.03846/) (2025)
- [Gradient Routing: Masking Gradients to Localize Computation in Neural Networks](https://mlsanity.com/paper/2410.04332/) (2024)
- [How Susceptible are LLMs to Influence in Prompts?](https://mlsanity.com/paper/2408.11865/) (2024)
- [Reliable Weak-to-Strong Monitoring of LLM Agents](https://mlsanity.com/paper/2508.19461/) (2025)
- [Reference-Guided Verdict: LLMs-as-Judges in Automatic Evaluation of Free-Form QA](https://mlsanity.com/paper/2408.09235/) (2024)
- [How AI Impacts Skill Formation](https://mlsanity.com/paper/2601.20245/) (2026)
- [Scaling Laws For Scalable Oversight](https://mlsanity.com/paper/2504.18530/) (2025)
- [Weak-for-Strong: Training Weak Meta-Agent to Harness Strong Executors](https://mlsanity.com/paper/2504.04785/) (2025)
- [Sentient Agent as a Judge: Evaluating Higher-Order Social Cognition in Large Language Models](https://mlsanity.com/paper/2505.02847/) (2025)
- [Multiple LLM Agents Debate for Equitable Cultural Alignment](https://mlsanity.com/paper/2505.24671/) (2025)
- [Multi-Agent Deep Research: Training Multi-Agent Systems with M-GRPO](https://mlsanity.com/paper/2511.13288/) (2025)
- [Training Language Models to Win Debates with Self-Play Improves Judge Accuracy](https://mlsanity.com/paper/2409.16636/) (2024)
- [An alignment safety case sketch based on debate](https://mlsanity.com/paper/2505.03989/) (2025)
- [Privacy Leakage Overshadowed by Views of AI: A Study on Human Oversight of Privacy in Language Model Agent](https://mlsanity.com/paper/2411.01344/) (2024)
- [Avoiding Obfuscation with Prover-Estimator Debate](https://mlsanity.com/paper/2506.13609/) (2025)
- [The Road to Artificial SuperIntelligence: A Comprehensive Survey of Superalignment](https://mlsanity.com/paper/2412.16468/) (2024)
- [When Persuasion Overrides Truth in Multi-Agent LLM Debates: Introducing a Confidence-Weighted Persuasion Override Rate (CW-POR)](https://mlsanity.com/paper/2504.00374/) (2025)
- [Distributional AGI Safety](https://mlsanity.com/paper/2512.16856/) (2025)
- [A Simple "Try Again" Can Elicit Multi-Turn LLM Reasoning](https://mlsanity.com/paper/2507.14295/) (2025)
- [Soteria: Language-Specific Functional Parameter Steering for Multilingual Safety Alignment](https://mlsanity.com/paper/2502.11244/) (2025)
- [A Survey of Scaling in Large Language Model Reasoning](https://mlsanity.com/paper/2504.02181/) (2025)
- [Control Tax: The Price of Keeping AI in Check](https://mlsanity.com/paper/2506.05296/) (2025)
- [More Convincing, Not More Correct: Self-Play Reward Hacking of Reference-Free LLM Judges](https://mlsanity.com/paper/2607.05904/) (2026)
- [Adaptive Collaboration with Humans: Metacognitive Policy Optimization for Multi-Agent LLMs with Continual Learning](https://mlsanity.com/paper/2603.07972/) (2026)

## Abstract (arXiv)

Scalable oversight protocols aim to enable humans to accurately supervise superhuman AI. In this paper we study debate, where two AI's compete to convince a judge; consultancy, where a single AI tries to convince a judge that asks questions; and compare to a baseline of direct question-answering, where the judge just answers outright without the AI. We use large language models (LLMs) as both AI agents and as stand-ins for human judges, taking the judge models to be weaker than agent models. We benchmark on a diverse range of asymmetries between judges and agents, extending previous work on a single extractive QA task with information asymmetry, to also include mathematics, coding, logic and multimodal reasoning asymmetries. We find that debate outperforms consultancy across all tasks when the consultant is randomly assigned to argue for the correct/incorrect answer. Comparing debate to direct question answering, the results depend on the type of task: in extractive QA tasks with information asymmetry debate outperforms direct question answering, but in other tasks without information asymmetry the results are mixed. Previous work assigned debaters/consultants an answer to argue for. When we allow them to instead choose which answer to argue for, we find judges are less frequently convinced by the wrong answer in debate than in consultancy. Further, we find that stronger debater models increase judge accuracy, though more modestly than in previous studies.