# Natural Language Understanding on NLU Benchmarks (CSQA, HellaSwag, PIQA, SIQA, ARC Easy) 5-shot Source: [Natural Language Understanding on NLU Benchmarks (CSQA, HellaSwag, PIQA, SIQA, ARC Easy) 5-shot]() Collection index: [llms.txt]() Data updated: 2026-07-09T07:52:02.192Z Evaluation of general Natural Language Understanding capabilities using a 5-shot prompting strategy across the CSQA, HellaSwag, PIQA, SIQA, and ARC Easy benchmarks, reporting the Core Average score. 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 | Natural Language Understanding | | Dataset (including any stated split) | NLU Benchmarks (CSQA, HellaSwag, PIQA, SIQA, ARC Easy) 5-shot | | Metric | CSQA Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 12 of 12 | | 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 - [CSQA Accuracy](): Higher is better; unit: PERCENT. - [HellaSwag Accuracy](): Higher is better; unit: PERCENT. - [PIQA Accuracy](): Higher is better; unit: PERCENT. - [SIQA Accuracy](): Higher is better; unit: PERCENT. - [ARC Easy Accuracy](): Higher is better; unit: PERCENT. - [Core Average Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | CSQA Accuracy | Result date | Report paper | | --- | --- | --- | --- | | HF/SMOLLM2-1.7B\* | 66.2 | Not recorded | 6a4f2b3b32fb937314875e42 | | HF/SMOLLM2-360M\* | 57.8 | Not recorded | 6a4f2b3b32fb937314875e42 | | CO-LMLM-360M-FW | 48.3 | Not recorded | 6a4f2b3b32fb937314875e42 | | HF/SMOLLM2-135M\* | 46.3 | Not recorded | 6a4f2b3b32fb937314875e42 | | STANDARD-360M-FW | 46 | Not recorded | 6a4f2b3b32fb937314875e42 | | STANDARD-360M | 32 | Not recorded | 6a4f2b3b32fb937314875e42 | | REL-LMLM-360M | 31.7 | Not recorded | 6a4f2b3b32fb937314875e42 | | REL-LMLM-135M | 29.9 | Not recorded | 6a4f2b3b32fb937314875e42 | | CO-LMLM-360M | 29.4 | Not recorded | 6a4f2b3b32fb937314875e42 | | STANDARD-135M | 27.1 | Not recorded | 6a4f2b3b32fb937314875e42 | | CO-LMLM-135M | 26.7 | Not recorded | 6a4f2b3b32fb937314875e42 | | Random Chance | 20 | Not recorded | 6a4f2b3b32fb937314875e42 | ### HF/SMOLLM2-1.7B\* CSQA Accuracy: 66.2. Result ID: 6a4f531460b88bb339e5511e. [Reporting paper]() Recorded evaluation configuration: {"backbone\_size":"1.7B","pre-training\_corpus":"2T-11T tokens","evaluation\_protocol":"5-shot"}. This record does not assert independent reproduction. ### HF/SMOLLM2-360M\* CSQA Accuracy: 57.8. Result ID: 6a4f531460b88bb339e5511d. [Reporting paper]() Recorded evaluation configuration: {"backbone\_size":"360M","pre-training\_corpus":"2T-11T tokens","evaluation\_protocol":"5-shot"}. This record does not assert independent reproduction. ### CO-LMLM-360M-FW CSQA Accuracy: 48.3. Result ID: 6a4f531460b88bb339e5511b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"backbone\_size":"360M","pre-training\_corpus":"Fewer than 100B tokens","evaluation\_protocol":"5-shot"}. This record does not assert independent reproduction. ### HF/SMOLLM2-135M\* CSQA Accuracy: 46.3. Result ID: 6a4f531460b88bb339e5511c. [Reporting paper]() Recorded evaluation configuration: {"backbone\_size":"135M","pre-training\_corpus":"2T-11T tokens","evaluation\_protocol":"5-shot"}. This record does not assert independent reproduction. ### STANDARD-360M-FW CSQA Accuracy: 46. Result ID: 6a4f531460b88bb339e5511a. [Reporting paper]() Recorded evaluation configuration: {"backbone\_size":"360M","pre-training\_corpus":"Fewer than 100B tokens","evaluation\_protocol":"5-shot"}. This record does not assert independent reproduction. ### STANDARD-360M CSQA Accuracy: 32. Result ID: 6a4f531460b88bb339e55118. [Reporting paper]() Recorded evaluation configuration: {"backbone\_size":"360M","pre-training\_corpus":"Wikipedia","evaluation\_protocol":"5-shot"}. This record does not assert independent reproduction. ### REL-LMLM-360M CSQA Accuracy: 31.7. Result ID: 6a4f531460b88bb339e55117. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"backbone\_size":"360M","evaluation\_protocol":"5-shot"}. This record does not assert independent reproduction. ### REL-LMLM-135M CSQA Accuracy: 29.9. Result ID: 6a4f531460b88bb339e55114. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"backbone\_size":"135M","evaluation\_protocol":"5-shot"}. This record does not assert independent reproduction. ### CO-LMLM-360M CSQA Accuracy: 29.4. Result ID: 6a4f531460b88bb339e55119. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"backbone\_size":"360M","pre-training\_corpus":"Wikipedia","evaluation\_protocol":"5-shot"}. This record does not assert independent reproduction. ### STANDARD-135M CSQA Accuracy: 27.1. Result ID: 6a4f531460b88bb339e55115. [Reporting paper]() Recorded evaluation configuration: {"backbone\_size":"135M","pre-training\_corpus":"Wikipedia","evaluation\_protocol":"5-shot"}. This record does not assert independent reproduction. ### CO-LMLM-135M CSQA Accuracy: 26.7. Result ID: 6a4f531460b88bb339e55116. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"backbone\_size":"135M","pre-training\_corpus":"Wikipedia","evaluation\_protocol":"5-shot"}. This record does not assert independent reproduction. ### Random Chance CSQA Accuracy: 20. Result ID: 6a4f531460b88bb339e55113. [Reporting paper]() Recorded evaluation configuration: {"evaluation\_protocol":"5-shot"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.