# Reasoning and Question Answering on BoolQ, RTE, HellaSWAG, ARC, OpenBookQA, and PiQA

Source: [Reasoning and Question Answering on BoolQ, RTE, HellaSWAG, ARC, OpenBookQA, and PiQA](<https://www.sota2.com/research/sota/reasoning-and-question-answering-on-boolq-rte-hellaswag-arc-openbookqa-and-piqa>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T05:36:04.320Z

Evaluation suite covering reasoning and question answering across the standard LLM benchmarks: BoolQ, RTE, HellaSWAG, ARC, OpenBookQA, and PiQA.

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 | Reasoning and Question Answering |
| Dataset (including any stated split) | Standard LLM Benchmarks (BoolQ, RTE, HellaSWAG, ARC, OpenBookQA, PiQA) |
| Metric | Avg Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 15 of 15 |
| 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

- [Avg Accuracy](<https://www.sota2.com/research/sota/reasoning-and-question-answering-on-boolq-rte-hellaswag-arc-openbookqa-and-piqa.md?metric=Avg%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Avg Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Before finetune | 67.24 | Not recorded | 68fa73941e209131cea5d553 |
| Target LLM | 66.93 | Not recorded | 68fa73941e209131cea5d553 |
| ULD | 66.85 | Not recorded | 68fa73941e209131cea5d553 |
| NPO+GD | 61.77 | Not recorded | 68fa73941e209131cea5d553 |
| NPO+KL | 61.14 | Not recorded | 68fa73941e209131cea5d553 |
| Offset-NPO+KL | 58.72 | Not recorded | 68fa73941e209131cea5d553 |
| GA+GD | 58.34 | Not recorded | 68fa73941e209131cea5d553 |
| Offset-DPO+KL | 56.59 | Not recorded | 68fa73941e209131cea5d553 |
| DPO+KL | 56.34 | Not recorded | 68fa73941e209131cea5d553 |
| GA+KL | 55.41 | Not recorded | 68fa73941e209131cea5d553 |
| NPO | 54.73 | Not recorded | 68fa73941e209131cea5d553 |
| DPO+GD | 53.91 | Not recorded | 68fa73941e209131cea5d553 |
| Offset-GA+KL | 53.78 | Not recorded | 68fa73941e209131cea5d553 |
| DPO | 48.12 | Not recorded | 68fa73941e209131cea5d553 |
| GA | 35.59 | Not recorded | 68fa73941e209131cea5d553 |

### Before finetune

Avg Accuracy: 67.24. Result ID: 6992f3ac3a6c314849129bd3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### Target LLM

Avg Accuracy: 66.93. Result ID: 6992f3ac3a6c314849129bd2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71c71e209131cea52ea2.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### ULD

Avg Accuracy: 66.85. Result ID: 6992f3ac3a6c314849129be0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Implementation](<https://github.com/UCSB-NLP-Chang/ULD>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### NPO+GD

Avg Accuracy: 61.77. Result ID: 6992f3ac3a6c314849129bdb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa753a1e209131cea66964.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### NPO+KL

Avg Accuracy: 61.14. Result ID: 6992f3ac3a6c314849129bdc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa753a1e209131cea66964.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### Offset-NPO+KL

Avg Accuracy: 58.72. Result ID: 6992f3ac3a6c314849129bdf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78d21e209131cea7a978.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### GA+GD

Avg Accuracy: 58.34. Result ID: 6992f3ac3a6c314849129bd5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71c71e209131cea52ea2.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### Offset-DPO+KL

Avg Accuracy: 56.59. Result ID: 6992f3ac3a6c314849129bde.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78d21e209131cea7a978.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### DPO+KL

Avg Accuracy: 56.34. Result ID: 6992f3ac3a6c314849129bd9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa753a1e209131cea66964.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### GA+KL

Avg Accuracy: 55.41. Result ID: 6992f3ac3a6c314849129bd6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa753a1e209131cea66964.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### NPO

Avg Accuracy: 54.73. Result ID: 6992f3ac3a6c314849129bda.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa753a1e209131cea66964.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### DPO+GD

Avg Accuracy: 53.91. Result ID: 6992f3ac3a6c314849129bd8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71c71e209131cea52ea2.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### Offset-GA+KL

Avg Accuracy: 53.78. Result ID: 6992f3ac3a6c314849129bdd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78d21e209131cea7a978.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### DPO

Avg Accuracy: 48.12. Result ID: 6992f3ac3a6c314849129bd7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71c71e209131cea52ea2.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

### GA

Avg Accuracy: 35.59. Result ID: 6992f3ac3a6c314849129bd4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73941e209131cea5d553.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71c71e209131cea52ea2.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-instruct","Mode":"Zero-shot"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
