# Verification on VerifyBench 1.0 (test) Source: [Verification on VerifyBench 1.0 (test)]() Collection index: [llms.txt]() Data updated: 2026-02-26T06:09:19.586Z Evaluation of model performance on the Verification task using the test split of the VerifyBench 1.0 dataset, measured by mean@3 accuracy. 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 | Verification | | Dataset (including any stated split) | VerifyBench 1.0 (test) | | Metric | m@3 Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 18 of 18 | | 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 - [m@3 Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | m@3 Accuracy | Result date | Report paper | | --- | --- | --- | --- | | CoSineVerifier-Tool-4B | 96.6 | Not recorded | 692e72fbd6b0dd3342932618 | | CompassVerifier-32B | 96.3 | Not recorded | 692e72fbd6b0dd3342932618 | | o3 | 96.1 | Not recorded | 692e72fbd6b0dd3342932618 | | GPT-4o | 96 | Not recorded | 692e72fbd6b0dd3342932618 | | Gemini2.5-Flash | 96 | Not recorded | 692e72fbd6b0dd3342932618 | | CompassVerifier-32B (CoT) | 95.9 | Not recorded | 692e72fbd6b0dd3342932618 | | CoSineVerifier-4B | 95.7 | Not recorded | 692e72fbd6b0dd3342932618 | | CoSineVerifier-32B | 95.7 | Not recorded | 692e72fbd6b0dd3342932618 | | LLaMA3.3-70B-Instruct | 94.8 | Not recorded | 692e72fbd6b0dd3342932618 | | Qwen3-32B | 94.7 | Not recorded | 692e72fbd6b0dd3342932618 | | Qwen3-4B-Instruct-2507 | 94.7 | Not recorded | 692e72fbd6b0dd3342932618 | | Qwen3-235B-A22B-2507 | 94.4 | Not recorded | 692e72fbd6b0dd3342932618 | | Qwen3-8B | 93.7 | Not recorded | 692e72fbd6b0dd3342932618 | | CompassVerifier-7B (CoT) | 93.5 | Not recorded | 692e72fbd6b0dd3342932618 | | CompassVerifier-7B | 93.5 | Not recorded | 692e72fbd6b0dd3342932618 | | Qwen3-4B | 92.6 | Not recorded | 692e72fbd6b0dd3342932618 | | XVerify-8B-I | 92.5 | Not recorded | 692e72fbd6b0dd3342932618 | | GPT-oss-20B | 92.2 | Not recorded | 692e72fbd6b0dd3342932618 | ### CoSineVerifier-Tool-4B m@3 Accuracy: 96.6. Result ID: 699998be5698665ec40d158a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"CoT Verifier","Model Access Category":"Open-source","Avg. Tokens":95.3}. This record does not assert independent reproduction. ### CompassVerifier-32B m@3 Accuracy: 96.3. Result ID: 699998be5698665ec40d158d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"Labeling Verifier","Avg. Tokens":1}. This record does not assert independent reproduction. ### o3 m@3 Accuracy: 96.1. Result ID: 699998be5698665ec40d157e. [Reporting paper]() Recorded evaluation configuration: {"Verifier Type":"CoT Verifier","Model Access Category":"Closed-source","Avg. Tokens":206.7}. This record does not assert independent reproduction. ### GPT-4o m@3 Accuracy: 96. Result ID: 699998be5698665ec40d157f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"CoT Verifier","Model Access Category":"Closed-source","Avg. Tokens":192.4}. This record does not assert independent reproduction. ### Gemini2.5-Flash m@3 Accuracy: 96. Result ID: 699998be5698665ec40d1580. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"CoT Verifier","Model Access Category":"Closed-source","Avg. Tokens":193}. This record does not assert independent reproduction. ### CompassVerifier-32B (CoT) m@3 Accuracy: 95.9. Result ID: 699998be5698665ec40d1589. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"CoT Verifier","Model Access Category":"Open-source","Avg. Tokens":213}. This record does not assert independent reproduction. ### CoSineVerifier-4B m@3 Accuracy: 95.7. Result ID: 699998be5698665ec40d158e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"Labeling Verifier","Avg. Tokens":1}. This record does not assert independent reproduction. ### CoSineVerifier-32B m@3 Accuracy: 95.7. Result ID: 699998be5698665ec40d158f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"Labeling Verifier","Avg. Tokens":1}. This record does not assert independent reproduction. ### LLaMA3.3-70B-Instruct m@3 Accuracy: 94.8. Result ID: 699998be5698665ec40d1582. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Verifier Type":"CoT Verifier","Model Access Category":"Open-source","Avg. Tokens":347.3}. This record does not assert independent reproduction. ### Qwen3-32B m@3 Accuracy: 94.7. Result ID: 699998be5698665ec40d1585. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"CoT Verifier","Model Access Category":"Open-source","Avg. Tokens":798.8}. This record does not assert independent reproduction. ### Qwen3-4B-Instruct-2507 m@3 Accuracy: 94.7. Result ID: 699998be5698665ec40d1586. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"CoT Verifier","Model Access Category":"Open-source","Avg. Tokens":869.7}. This record does not assert independent reproduction. ### Qwen3-235B-A22B-2507 m@3 Accuracy: 94.4. Result ID: 699998be5698665ec40d1587. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"CoT Verifier","Model Access Category":"Open-source","Avg. Tokens":1885.3}. This record does not assert independent reproduction. ### Qwen3-8B m@3 Accuracy: 93.7. Result ID: 699998be5698665ec40d1584. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"CoT Verifier","Model Access Category":"Open-source","Avg. Tokens":926.6}. This record does not assert independent reproduction. ### CompassVerifier-7B (CoT) m@3 Accuracy: 93.5. Result ID: 699998be5698665ec40d1588. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"CoT Verifier","Model Access Category":"Open-source","Avg. Tokens":234.7}. This record does not assert independent reproduction. ### CompassVerifier-7B m@3 Accuracy: 93.5. Result ID: 699998be5698665ec40d158c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"Labeling Verifier","Avg. Tokens":1}. This record does not assert independent reproduction. ### Qwen3-4B m@3 Accuracy: 92.6. Result ID: 699998be5698665ec40d1583. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"CoT Verifier","Model Access Category":"Open-source","Avg. Tokens":1156.7}. This record does not assert independent reproduction. ### XVerify-8B-I m@3 Accuracy: 92.5. Result ID: 699998be5698665ec40d158b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"Labeling Verifier","Avg. Tokens":1}. This record does not assert independent reproduction. ### GPT-oss-20B m@3 Accuracy: 92.2. Result ID: 699998be5698665ec40d1581. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Verifier Type":"CoT Verifier","Model Access Category":"Open-source","Avg. Tokens":221}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.