# Question Answering on MathQA, MMLU, HellaSwag, RACE, and CosmosQA Source: [Question Answering on MathQA, MMLU, HellaSwag, RACE, and CosmosQA]() Collection index: [llms.txt]() Data updated: 2026-04-08T06:36:53.142Z Evaluation of model performance across five distinct Question Answering datasets: MathQA, MMLU, HellaSwag, RACE, and CosmosQA, reporting accuracy and calibration metrics (ECE, ACE, Brier). 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 | Question Answering | | Dataset (including any stated split) | MathQA, MMLU, HellaSwag, RACE, and CosmosQA | | Metric | ECE | | Unit | SCALAR | | Direction | Lower is better | | Results loaded | 27 of 27 | | 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 - [ECE](): Lower is better; unit: SCALAR. - [ACE](): Lower is better; unit: SCALAR. - [Brier Score](): Lower is better; unit: SCALAR. - [Accuracy](): Higher is better; unit: PERCENT. - [AUROC](): Higher is better; unit: PERCENT. ## Recorded results | Method | ECE | Result date | Report paper | | --- | --- | --- | --- | | R1 Llama | 0.079 | Not recorded | 69d5e1087d0172140ec75a78 | | Llama 3.3 | 0.121 | Not recorded | 69d5e1087d0172140ec75a78 | | Phi-4 mini | 0.128 | Not recorded | 69d5e1087d0172140ec75a78 | | Phi-4 mini | 0.132 | Not recorded | 69d5e1087d0172140ec75a78 | | Llama 3.3 | 0.146 | Not recorded | 69d5e1087d0172140ec75a78 | | Llama 3.3 | 0.148 | Not recorded | 69d5e1087d0172140ec75a78 | | R1 Qwen3 | 0.166 | Not recorded | 69d5e1087d0172140ec75a78 | | R1 Qwen3 | 0.167 | Not recorded | 69d5e1087d0172140ec75a78 | | GPT-oss | 0.191 | Not recorded | 69d5e1087d0172140ec75a78 | | GPT-oss | 0.191 | Not recorded | 69d5e1087d0172140ec75a78 | | Phi-4 | 0.194 | Not recorded | 69d5e1087d0172140ec75a78 | | GPT-oss | 0.197 | Not recorded | 69d5e1087d0172140ec75a78 | | R1 Qwen3 | 0.198 | Not recorded | 69d5e1087d0172140ec75a78 | | Phi-4 | 0.199 | Not recorded | 69d5e1087d0172140ec75a78 | | Qwen 2.5 | 0.205 | Not recorded | 69d5e1087d0172140ec75a78 | | Phi-4 | 0.209 | Not recorded | 69d5e1087d0172140ec75a78 | | R1 Llama | 0.245 | Not recorded | 69d5e1087d0172140ec75a78 | | Qwen 2.5 | 0.245 | Not recorded | 69d5e1087d0172140ec75a78 | | R1 Llama | 0.267 | Not recorded | 69d5e1087d0172140ec75a78 | | Phi-4 mini | 0.285 | Not recorded | 69d5e1087d0172140ec75a78 | | R1 Qwen | 0.312 | Not recorded | 69d5e1087d0172140ec75a78 | | R1 Qwen | 0.318 | Not recorded | 69d5e1087d0172140ec75a78 | | Qwen 2.5 | 0.329 | Not recorded | 69d5e1087d0172140ec75a78 | | R1 Qwen | 0.334 | Not recorded | 69d5e1087d0172140ec75a78 | | Nemotron | 0.334 | Not recorded | 69d5e1087d0172140ec75a78 | | Nemotron | 0.409 | Not recorded | 69d5e1087d0172140ec75a78 | | Nemotron | 0.415 | Not recorded | 69d5e1087d0172140ec75a78 | ### R1 Llama ECE: 0.079. Result ID: 69d5f6fe47f34ce924bdf910. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Suppression","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### Llama 3.3 ECE: 0.121. Result ID: 69d5f6fe47f34ce924bdf91e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"Baseline Inference","Model Category":"Instruction-tuned Model"}. This record does not assert independent reproduction. ### Phi-4 mini ECE: 0.128. Result ID: 69d5f6fe47f34ce924bdf925. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Suppression","Model Category":"Instruction-tuned Model"}. This record does not assert independent reproduction. ### Phi-4 mini ECE: 0.132. Result ID: 69d5f6fe47f34ce924bdf926. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Injection","Model Category":"Instruction-tuned Model"}. This record does not assert independent reproduction. ### Llama 3.3 ECE: 0.146. Result ID: 69d5f6fe47f34ce924bdf91f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Suppression","Model Category":"Instruction-tuned Model"}. This record does not assert independent reproduction. ### Llama 3.3 ECE: 0.148. Result ID: 69d5f6fe47f34ce924bdf920. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Injection","Model Category":"Instruction-tuned Model"}. This record does not assert independent reproduction. ### R1 Qwen3 ECE: 0.166. Result ID: 69d5f6fe47f34ce924bdf913. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Suppression","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### R1 Qwen3 ECE: 0.167. Result ID: 69d5f6fe47f34ce924bdf914. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Injection","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### GPT-oss ECE: 0.191. Result ID: 69d5f6fe47f34ce924bdf91b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"Baseline Inference","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### GPT-oss ECE: 0.191. Result ID: 69d5f6fe47f34ce924bdf91c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Suppression","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### Phi-4 ECE: 0.194. Result ID: 69d5f6fe47f34ce924bdf919. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Suppression","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### GPT-oss ECE: 0.197. Result ID: 69d5f6fe47f34ce924bdf91d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Injection","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### R1 Qwen3 ECE: 0.198. Result ID: 69d5f6fe47f34ce924bdf912. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"Baseline Inference","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### Phi-4 ECE: 0.199. Result ID: 69d5f6fe47f34ce924bdf918. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"Baseline Inference","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### Qwen 2.5 ECE: 0.205. Result ID: 69d5f6fe47f34ce924bdf922. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Suppression","Model Category":"Instruction-tuned Model"}. This record does not assert independent reproduction. ### Phi-4 ECE: 0.209. Result ID: 69d5f6fe47f34ce924bdf91a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Injection","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### R1 Llama ECE: 0.245. Result ID: 69d5f6fe47f34ce924bdf90f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"Baseline Inference","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### Qwen 2.5 ECE: 0.245. Result ID: 69d5f6fe47f34ce924bdf923. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Injection","Model Category":"Instruction-tuned Model"}. This record does not assert independent reproduction. ### R1 Llama ECE: 0.267. Result ID: 69d5f6fe47f34ce924bdf911. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Injection","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### Phi-4 mini ECE: 0.285. Result ID: 69d5f6fe47f34ce924bdf924. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"Baseline Inference","Model Category":"Instruction-tuned Model"}. This record does not assert independent reproduction. ### R1 Qwen ECE: 0.312. Result ID: 69d5f6fe47f34ce924bdf90e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Injection","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### R1 Qwen ECE: 0.318. Result ID: 69d5f6fe47f34ce924bdf90d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Suppression","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### Qwen 2.5 ECE: 0.329. Result ID: 69d5f6fe47f34ce924bdf921. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"Baseline Inference","Model Category":"Instruction-tuned Model"}. This record does not assert independent reproduction. ### R1 Qwen ECE: 0.334. Result ID: 69d5f6fe47f34ce924bdf90c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"Baseline Inference","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### Nemotron ECE: 0.334. Result ID: 69d5f6fe47f34ce924bdf916. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Suppression","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### Nemotron ECE: 0.409. Result ID: 69d5f6fe47f34ce924bdf917. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"N-gram Injection","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. ### Nemotron ECE: 0.415. Result ID: 69d5f6fe47f34ce924bdf915. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"Baseline Inference","Model Category":"Reasoning Model"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.