# Science Question Answering on ARC-C (Accuracy and Gap) Source: [Science Question Answering on ARC-C (Accuracy and Gap)]() Collection index: [llms.txt]() Data updated: 2026-03-10T08:08:44.389Z Evaluation of model performance on the ARC-C dataset for Science Question Answering, reporting Accuracy and the Gap metric. 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 | Science Question Answering | | Dataset (including any stated split) | ARC-C | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 32 of 32 | | 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 - [Accuracy](): Higher is better; unit: PERCENT. - [Gap](): Lower is better; unit: SCALAR. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | Evo 8B | 92.5 | Not recorded | 69afa57c2d603e6de3e439b5 | | LLaDA 8B | 88.5 | Not recorded | 69afa57c2d603e6de3e439b5 | | Qwen2.5-Omni-7B | 87.1 | Not recorded | 68fa7c2c1e209131cea8cdd9 | | BD3-LM 7B | 86.7 | Not recorded | 69afa57c2d603e6de3e439b5 | | ASR + Qwen2.5-7B | 86.5 | Not recorded | 68fa7c2c1e209131cea8cdd9 | | SALAD-7B | 84 | Not recorded | 68fa7c2c1e209131cea8cdd9 | | AceMAD | 82.43 | Not recorded | 69afa5822d603e6de3e43a6d | | LLaMA3 8B | 82.4 | Not recorded | 69afa57c2d603e6de3e439b5 | | SALAD-7B | 82.3 | Not recorded | 68fa7c2c1e209131cea8cdd9 | | AceMAD | 80.56 | Not recorded | 69afa5822d603e6de3e43a6d | | SALAD-3B | 79.9 | Not recorded | 68fa7c2c1e209131cea8cdd9 | | AceMAD | 77.78 | Not recorded | 69afa5822d603e6de3e43a6d | | ASR + Qwen2.5-3B | 77.7 | Not recorded | 68fa7c2c1e209131cea8cdd9 | | SALAD-3B | 75.6 | Not recorded | 68fa7c2c1e209131cea8cdd9 | | Majority Voting | 69.44 | Not recorded | 69afa5822d603e6de3e43a6d | | Single Agent | 65.74 | Not recorded | 69afa5822d603e6de3e43a6d | | GLM-4-Voice-9B | 64.6 | Not recorded | 68fa7c2c1e209131cea8cdd9 | | AceMAD | 59.26 | Not recorded | 69afa5822d603e6de3e43a6d | | ARD 7B | 56.9 | Not recorded | 69afa57c2d603e6de3e439b5 | | AceMAD | 56.48 | Not recorded | 69afa5822d603e6de3e43a6d | | AceMAD | 56.48 | Not recorded | 69afa5822d603e6de3e43a6d | | Decentralized MAD | 49.07 | Not recorded | 69afa5822d603e6de3e43a6d | | DiVA-Llama3.1-8B | 45.9 | Not recorded | 68fa7c2c1e209131cea8cdd9 | | Sparse MAD | 43.52 | Not recorded | 69afa5822d603e6de3e43a6d | | Qwen2-Audio-7B | 43.5 | Not recorded | 68fa7c2c1e209131cea8cdd9 | | Decentralized MAD | 41.67 | Not recorded | 69afa5822d603e6de3e43a6d | | Centralized MAD | 41.67 | Not recorded | 69afa5822d603e6de3e43a6d | | Random | 25 | Not recorded | 68fa7c2c1e209131cea8cdd9 | | Majority Voting | 23.15 | Not recorded | 69afa5822d603e6de3e43a6d | | Single Agent | 20.37 | Not recorded | 69afa5822d603e6de3e43a6d | | Centralized MAD | 13.89 | Not recorded | 69afa5822d603e6de3e43a6d | | Sparse MAD | 13.89 | Not recorded | 69afa5822d603e6de3e43a6d | ### Evo 8B Accuracy: 92.5. Result ID: 69afcbfd6dbb7075b957dc0b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Post-training":"SFT","Number of shots":"0"}. This record does not assert independent reproduction. ### LLaDA 8B Accuracy: 88.5. Result ID: 69afcbfd6dbb7075b957dc0e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Post-training":"SFT","Number of shots":"0"}. This record does not assert independent reproduction. ### Qwen2.5-Omni-7B Accuracy: 87.1. Result ID: 699d4de75698665ec4f032e8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"System Type":"End-to-end system"}. This record does not assert independent reproduction. ### BD3-LM 7B Accuracy: 86.7. Result ID: 69afcbfd6dbb7075b957dc0c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Post-training":"SFT+RL","Number of shots":"0"}. This record does not assert independent reproduction. ### ASR + Qwen2.5-7B Accuracy: 86.5. Result ID: 699d4de75698665ec4f032e4. [Reporting paper]() Recorded evaluation configuration: {"System Type":"Cascaded topline","ASR":"Whisper-v3-Large"}. This record does not assert independent reproduction. ### SALAD-7B Accuracy: 84. Result ID: 699d4de75698665ec4f032ec. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training Stage":"Stage II"}. This record does not assert independent reproduction. ### AceMAD Accuracy: 82.43. Result ID: 69afd18b6dbb7075b95805d9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"Qwen3-235B-A22B-Instruct","Iteration Rounds (T)":"2","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### LLaMA3 8B Accuracy: 82.4. Result ID: 69afcbfd6dbb7075b957dc0f. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Post-training":"SFT+RL","Number of shots":"0"}. This record does not assert independent reproduction. ### SALAD-7B Accuracy: 82.3. Result ID: 699d4de75698665ec4f032eb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training Stage":"Stage I"}. This record does not assert independent reproduction. ### AceMAD Accuracy: 80.56. Result ID: 69afd18b6dbb7075b95805da. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"Qwen3-235B-A22B-Instruct","Iteration Rounds (T)":"3","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### SALAD-3B Accuracy: 79.9. Result ID: 699d4de75698665ec4f032ea. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training Stage":"Stage II"}. This record does not assert independent reproduction. ### AceMAD Accuracy: 77.78. Result ID: 69afd18b6dbb7075b95805db. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"Qwen3-235B-A22B-Instruct","Iteration Rounds (T)":"5","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### ASR + Qwen2.5-3B Accuracy: 77.7. Result ID: 699d4de75698665ec4f032e3. [Reporting paper]() Recorded evaluation configuration: {"System Type":"Cascaded topline","ASR":"Whisper-v3-Large"}. This record does not assert independent reproduction. ### SALAD-3B Accuracy: 75.6. Result ID: 699d4de75698665ec4f032e9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training Stage":"Stage I"}. This record does not assert independent reproduction. ### Majority Voting Accuracy: 69.44. Result ID: 69afd18b6dbb7075b95805d5. [Reporting paper]() Recorded evaluation configuration: {"Base Model":"Qwen3-235B-A22B-Instruct","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### Single Agent Accuracy: 65.74. Result ID: 69afd18b6dbb7075b95805d4. [Reporting paper]() Recorded evaluation configuration: {"Base Model":"Qwen3-235B-A22B-Instruct","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### GLM-4-Voice-9B Accuracy: 64.6. Result ID: 699d4de75698665ec4f032e7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"System Type":"End-to-end system"}. This record does not assert independent reproduction. ### AceMAD Accuracy: 59.26. Result ID: 69afd18b6dbb7075b95805d3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Iteration Rounds (T)":"5","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### ARD 7B Accuracy: 56.9. Result ID: 69afcbfd6dbb7075b957dc0d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Post-training":"SFT+RL","Number of shots":"0"}. This record does not assert independent reproduction. ### AceMAD Accuracy: 56.48. Result ID: 69afd18b6dbb7075b95805d1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Iteration Rounds (T)":"2","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### AceMAD Accuracy: 56.48. Result ID: 69afd18b6dbb7075b95805d2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Iteration Rounds (T)":"3","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### Decentralized MAD Accuracy: 49.07. Result ID: 69afd18b6dbb7075b95805d6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"Qwen3-235B-A22B-Instruct","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### DiVA-Llama3.1-8B Accuracy: 45.9. Result ID: 699d4de75698665ec4f032e6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"System Type":"End-to-end system"}. This record does not assert independent reproduction. ### Sparse MAD Accuracy: 43.52. Result ID: 69afd18b6dbb7075b95805d0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### Qwen2-Audio-7B Accuracy: 43.5. Result ID: 699d4de75698665ec4f032e5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"System Type":"End-to-end system"}. This record does not assert independent reproduction. ### Decentralized MAD Accuracy: 41.67. Result ID: 69afd18b6dbb7075b95805ce. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### Centralized MAD Accuracy: 41.67. Result ID: 69afd18b6dbb7075b95805cf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### Random Accuracy: 25. Result ID: 699d4de75698665ec4f032e2. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Majority Voting Accuracy: 23.15. Result ID: 69afd18b6dbb7075b95805cd. [Reporting paper]() Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### Single Agent Accuracy: 20.37. Result ID: 69afd18b6dbb7075b95805cc. [Reporting paper]() Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### Centralized MAD Accuracy: 13.89. Result ID: 69afd18b6dbb7075b95805d7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"Qwen3-235B-A22B-Instruct","Number of Agents (N)":"5"}. This record does not assert independent reproduction. ### Sparse MAD Accuracy: 13.89. Result ID: 69afd18b6dbb7075b95805d8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"Qwen3-235B-A22B-Instruct","Number of Agents (N)":"5"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.