# Step-level reasoning verification on Meetings Source: [Step-level reasoning verification on Meetings]() Collection index: [llms.txt]() Data updated: 2026-04-23T10:01:13.360Z Evaluation of model performance on the Step-level reasoning verification task using the Meetings dataset, measured by PR-AUC. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Step-level reasoning verification | | Dataset (including any stated split) | Meetings | | Metric | PR-AUC | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 19 of 19 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [PR-AUC](): Higher is better; unit: PERCENT. ## Recorded results | Method | PR-AUC | Result date | Report paper | | --- | --- | --- | --- | | ReProbe, Hidden States, GPT-OSS-anno | 73.4 | Not recorded | 6912c2ee2828b7cc6937d9ec | | Qwen2.5-Math-PRM-7B | 66 | Not recorded | 6912c2ee2828b7cc6937d9ec | | ReProbe, Attn+Logit, Qwen3-8B-anno | 64.6 | Not recorded | 6912c2ee2828b7cc6937d9ec | | Universal-PRM-Qwen2.5-Math-7B | 63.4 | Not recorded | 6912c2ee2828b7cc6937d9ec | | Qwen2.5-Math-7B-PRM800k | 59.9 | Not recorded | 6912c2ee2828b7cc6937d9ec | | Qwen2.5-Math-7B | 59.4 | Not recorded | 6912c2ee2828b7cc6937d9ec | | Qwen2.5-Math-7B-PRM800K | 59.1 | Not recorded | 6912c2ee2828b7cc6937d9ec | | Math-Shepherd-PRM-7B | 58.4 | Not recorded | 6912c2ee2828b7cc6937d9ec | | Skywork-PRM-1.5B | 56.5 | Not recorded | 6912c2ee2828b7cc6937d9ec | | Universal-PRM-Qwen2.5-Math-7B | 55.9 | Not recorded | 6912c2ee2828b7cc6937d9ec | | MaxProb | 54.8 | Not recorded | 6912c2ee2828b7cc6937d9ec | | MaxEntropy | 53.3 | Not recorded | 6912c2ee2828b7cc6937d9ec | | Perplexity | 50.8 | Not recorded | 6912c2ee2828b7cc6937d9ec | | H4-Qwen2.5-PRM-1.5B-0.2 | 47.6 | Not recorded | 6912c2ee2828b7cc6937d9ec | | Skywork-PRM-1.5B | 46.7 | Not recorded | 6912c2ee2828b7cc6937d9ec | | Random | 46.3 | Not recorded | 6912c2ee2828b7cc6937d9ec | | RLHFlow-PRM-Deepseek-8B | 45.5 | Not recorded | 6912c2ee2828b7cc6937d9ec | | RLHFlow-PRM-Mistral-8B | 42.4 | Not recorded | 6912c2ee2828b7cc6937d9ec | | H4-Qwen2.5-PRM-1.5B-0.2 | 32.5 | Not recorded | 6912c2ee2828b7cc6937d9ec | ### ReProbe, Hidden States, GPT-OSS-anno PR-AUC: 73.4. Result ID: 69e9edd747f34ce924d19e22. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"# Sample":"10.8K","Verifying Model":"Qwen3-32B","Thinking Mode":"native"}. This record does not assert independent reproduction. ### Qwen2.5-Math-PRM-7B PR-AUC: 66. Result ID: 69e9edd747f34ce924d19e21. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"# Sample":"860K","Verifying Model":"Qwen3-32B","Thinking Mode":"native"}. This record does not assert independent reproduction. ### ReProbe, Attn+Logit, Qwen3-8B-anno PR-AUC: 64.6. Result ID: 69e9edd747f34ce924d19ce0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"# Sample":"32K"}. This record does not assert independent reproduction. ### Universal-PRM-Qwen2.5-Math-7B PR-AUC: 63.4. Result ID: 69e9edd747f34ce924d19e1f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"# Sample":"690K","Verifying Model":"Qwen3-32B","Thinking Mode":"native"}. This record does not assert independent reproduction. ### Qwen2.5-Math-7B-PRM800k PR-AUC: 59.9. Result ID: 69e9edd747f34ce924d19cde. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"# Sample":"265K"}. This record does not assert independent reproduction. ### Qwen2.5-Math-7B PR-AUC: 59.4. Result ID: 69e9edd747f34ce924d19cdf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"# Sample":"860K"}. This record does not assert independent reproduction. ### Qwen2.5-Math-7B-PRM800K PR-AUC: 59.1. Result ID: 69e9edd747f34ce924d19e20. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"# Sample":"263K","Verifying Model":"Qwen3-32B","Thinking Mode":"native"}. This record does not assert independent reproduction. ### Math-Shepherd-PRM-7B PR-AUC: 58.4. Result ID: 69e9edd747f34ce924d19cda. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"# Sample":"440K"}. This record does not assert independent reproduction. ### Skywork-PRM-1.5B PR-AUC: 56.5. Result ID: 69e9edd747f34ce924d19e1d. [Reporting paper]() Recorded evaluation configuration: {"# Sample":"Unk","Verifying Model":"Qwen3-32B","Thinking Mode":"native"}. This record does not assert independent reproduction. ### Universal-PRM-Qwen2.5-Math-7B PR-AUC: 55.9. Result ID: 69e9edd747f34ce924d19cdd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"# Sample":"690K"}. This record does not assert independent reproduction. ### MaxProb PR-AUC: 54.8. Result ID: 69e9edd747f34ce924d19cd5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"# Sample":"-"}. This record does not assert independent reproduction. ### MaxEntropy PR-AUC: 53.3. Result ID: 69e9edd747f34ce924d19cd6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"# Sample":"-"}. This record does not assert independent reproduction. ### Perplexity PR-AUC: 50.8. Result ID: 69e9edd747f34ce924d19cd7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"# Sample":"-"}. This record does not assert independent reproduction. ### H4-Qwen2.5-PRM-1.5B-0.2 PR-AUC: 47.6. Result ID: 69e9edd747f34ce924d19cd9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"# Sample":"369K"}. This record does not assert independent reproduction. ### Skywork-PRM-1.5B PR-AUC: 46.7. Result ID: 69e9edd747f34ce924d19cd8. [Reporting paper]() Recorded evaluation configuration: {"# Sample":"Unk"}. This record does not assert independent reproduction. ### Random PR-AUC: 46.3. Result ID: 69e9edd747f34ce924d19cd4. [Reporting paper]() Recorded evaluation configuration: {"# Sample":"-"}. This record does not assert independent reproduction. ### RLHFlow-PRM-Deepseek-8B PR-AUC: 45.5. Result ID: 69e9edd747f34ce924d19cdb. [Reporting paper]() Recorded evaluation configuration: {"# Sample":"253K"}. This record does not assert independent reproduction. ### RLHFlow-PRM-Mistral-8B PR-AUC: 42.4. Result ID: 69e9edd747f34ce924d19cdc. [Reporting paper]() Recorded evaluation configuration: {"# Sample":"273K"}. This record does not assert independent reproduction. ### H4-Qwen2.5-PRM-1.5B-0.2 PR-AUC: 32.5. Result ID: 69e9edd747f34ce924d19e1e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"# Sample":"369K","Verifying Model":"Qwen3-32B","Thinking Mode":"native"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.