FindTheFlaws: Annotated Errors for Detecting Flawed Reasoning and Scalable Oversight Research
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866912300399591424 |
|---|---|
| author | Recchia, Gabriel Mangat, Chatrik Singh Li, Issac Krishnakumar, Gayatri |
| author_facet | Recchia, Gabriel Mangat, Chatrik Singh Li, Issac Krishnakumar, Gayatri |
| contents | As AI models tackle increasingly complex problems, ensuring reliable human oversight becomes more challenging due to the difficulty of verifying solutions. Approaches to scaling AI supervision include debate, in which two agents engage in structured dialogue to help a judge evaluate claims; critique, in which models identify potential flaws in proposed solutions; and prover-verifier games, in which a capable 'prover' model generates solutions that must be verifiable by a less capable 'verifier'. Evaluations of the scalability of these and similar approaches to difficult problems benefit from datasets that include (1) long-form expert-verified correct solutions and (2) long-form flawed solutions with annotations highlighting specific errors, but few are available.
To address this gap, we present FindTheFlaws, a group of five diverse datasets spanning medicine, mathematics, science, coding, and the Lojban language. Each dataset contains questions and long-form solutions with expert annotations validating their correctness or identifying specific error(s) in the reasoning. We evaluate frontier models' critiquing capabilities and observe a range of performance that can be leveraged for scalable oversight experiments: models performing more poorly on particular datasets can serve as judges/verifiers for more capable models. Additionally, for some task/dataset combinations, expert baselines exceed even top model performance, making them more beneficial for scalable oversight experiments. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_22989 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | FindTheFlaws: Annotated Errors for Detecting Flawed Reasoning and Scalable Oversight Research Recchia, Gabriel Mangat, Chatrik Singh Li, Issac Krishnakumar, Gayatri Artificial Intelligence Computation and Language I.2 As AI models tackle increasingly complex problems, ensuring reliable human oversight becomes more challenging due to the difficulty of verifying solutions. Approaches to scaling AI supervision include debate, in which two agents engage in structured dialogue to help a judge evaluate claims; critique, in which models identify potential flaws in proposed solutions; and prover-verifier games, in which a capable 'prover' model generates solutions that must be verifiable by a less capable 'verifier'. Evaluations of the scalability of these and similar approaches to difficult problems benefit from datasets that include (1) long-form expert-verified correct solutions and (2) long-form flawed solutions with annotations highlighting specific errors, but few are available. To address this gap, we present FindTheFlaws, a group of five diverse datasets spanning medicine, mathematics, science, coding, and the Lojban language. Each dataset contains questions and long-form solutions with expert annotations validating their correctness or identifying specific error(s) in the reasoning. We evaluate frontier models' critiquing capabilities and observe a range of performance that can be leveraged for scalable oversight experiments: models performing more poorly on particular datasets can serve as judges/verifiers for more capable models. Additionally, for some task/dataset combinations, expert baselines exceed even top model performance, making them more beneficial for scalable oversight experiments. |
| title | FindTheFlaws: Annotated Errors for Detecting Flawed Reasoning and Scalable Oversight Research |
| topic | Artificial Intelligence Computation and Language I.2 |
| url | https://arxiv.org/abs/2503.22989 |