# Zero-shot Language Understanding on ARC-C, BoolQ, Lambada, PIQA, and Winogrande

Source: [Zero-shot Language Understanding on ARC-C, BoolQ, Lambada, PIQA, and Winogrande](<https://www.sota2.com/research/sota/zero-shot-language-understanding-on-arc-c-boolq-lambada-piqa-and-winogrande>)

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

Data updated: 2026-03-31T09:10:41.432Z

Evaluation of zero-shot performance across the specific suite of language understanding tasks: ARC-Challenge, BoolQ, Lambada, PIQA, and Winogrande.

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 | Language Understanding |
| Dataset (including any stated split) | Benchmarks (ARC-C, BoolQ, Lambada, PIQA, Winogrande) Zero-shot |
| Metric | ARC-C Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 16 of 16 |
| 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

- [ARC-C Accuracy](<https://www.sota2.com/research/sota/zero-shot-language-understanding-on-arc-c-boolq-lambada-piqa-and-winogrande.md?metric=ARC-C%20Accuracy>): Higher is better; unit: PERCENT.

- [BoolQ Accuracy](<https://www.sota2.com/research/sota/zero-shot-language-understanding-on-arc-c-boolq-lambada-piqa-and-winogrande.md?metric=BoolQ%20Accuracy>): Higher is better; unit: PERCENT.

- [Lambada Accuracy](<https://www.sota2.com/research/sota/zero-shot-language-understanding-on-arc-c-boolq-lambada-piqa-and-winogrande.md?metric=Lambada%20Accuracy>): Higher is better; unit: PERCENT.

- [PIQA Accuracy](<https://www.sota2.com/research/sota/zero-shot-language-understanding-on-arc-c-boolq-lambada-piqa-and-winogrande.md?metric=PIQA%20Accuracy>): Higher is better; unit: PERCENT.

- [Winogrande Accuracy](<https://www.sota2.com/research/sota/zero-shot-language-understanding-on-arc-c-boolq-lambada-piqa-and-winogrande.md?metric=Winogrande%20Accuracy>): Higher is better; unit: PERCENT.

- [Average Score](<https://www.sota2.com/research/sota/zero-shot-language-understanding-on-arc-c-boolq-lambada-piqa-and-winogrande.md?metric=Average%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | ARC-C Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| MicroMix | 56.66 | Not recorded | 68fa7a6a1e209131cea83140 |
| FlatQuant | 56.23 | Not recorded | 68fa7a6a1e209131cea83140 |
| FP16 | 55.89 | Not recorded | 68fa7a6a1e209131cea83140 |
| INT6 | 55.29 | Not recorded | 68fa7a6a1e209131cea83140 |
| Atom | 54.78 | Not recorded | 68fa7a6a1e209131cea83140 |
| FP16 | 53.58 | Not recorded | 68fa7a6a1e209131cea83140 |
| QuaRot | 53.08 | Not recorded | 68fa7a6a1e209131cea83140 |
| QUIK | 52.9 | Not recorded | 68fa7a6a1e209131cea83140 |
| FlatQuant | 51.54 | Not recorded | 68fa7a6a1e209131cea83140 |
| AMXFP4 | 51.54 | Not recorded | 68fa7a6a1e209131cea83140 |
| MicroMix | 50.26 | Not recorded | 68fa7a6a1e209131cea83140 |
| Atom | 50.17 | Not recorded | 68fa7a6a1e209131cea83140 |
| INT6 | 48.38 | Not recorded | 68fa7a6a1e209131cea83140 |
| QuaRot | 46.42 | Not recorded | 68fa7a6a1e209131cea83140 |
| QUIK | 44.62 | Not recorded | 68fa7a6a1e209131cea83140 |
| AMXFP4 | 43.77 | Not recorded | 68fa7a6a1e209131cea83140 |

### MicroMix

ARC-C Accuracy: 56.66. Result ID: 69cb8e9347f34ce924b410d5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-32B","Avg. Bits":"5.22","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

### FlatQuant

ARC-C Accuracy: 56.23. Result ID: 69cb8e9347f34ce924b410d2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a8b1e209131cea83cda.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-32B","Avg. Bits":"4.71","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

### FP16

ARC-C Accuracy: 55.89. Result ID: 69cb8e9347f34ce924b410ce.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-32B","Avg. Bits":"16.00","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

### INT6

ARC-C Accuracy: 55.29. Result ID: 69cb8e9347f34ce924b410d4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-32B","Avg. Bits":"6.00","Evaluation Protocol":"0-shot","Quantization Implementation":"symmetric per-token"}. This record does not assert independent reproduction.

### Atom

ARC-C Accuracy: 54.78. Result ID: 69cb8e9347f34ce924b410d1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72f11e209131cea5970b.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-32B","Avg. Bits":"4.21","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

### FP16

ARC-C Accuracy: 53.58. Result ID: 69cb8e9347f34ce924b410c6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

Recorded evaluation configuration: {"Model":"Llama3.1-8B","Avg. Bits":"16.00","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

### QuaRot

ARC-C Accuracy: 53.08. Result ID: 69cb8e9347f34ce924b410cf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa756a1e209131cea67b8c.md>)

[Implementation](<https://github.com/spcl/QuaRot>)

Recorded evaluation configuration: {"Model":"Qwen2.5-32B","Avg. Bits":"4.12","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

### QUIK

ARC-C Accuracy: 52.9. Result ID: 69cb8e9347f34ce924b410d0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71141e209131cea4f205.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-32B","Avg. Bits":"6.21","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

### FlatQuant

ARC-C Accuracy: 51.54. Result ID: 69cb8e9347f34ce924b410ca.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a8b1e209131cea83cda.md>)

Recorded evaluation configuration: {"Model":"Llama3.1-8B","Avg. Bits":"4.19","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

### AMXFP4

ARC-C Accuracy: 51.54. Result ID: 69cb8e9347f34ce924b410d3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78f21e209131cea7b600.md>)

[Implementation](<https://www.sota2.com/aiha-lab/MX-QLLM.git>)

Recorded evaluation configuration: {"Model":"Qwen2.5-32B","Avg. Bits":"5.00","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

### MicroMix

ARC-C Accuracy: 50.26. Result ID: 69cb8e9347f34ce924b410cd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

Recorded evaluation configuration: {"Model":"Llama3.1-8B","Avg. Bits":"5.51","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

### Atom

ARC-C Accuracy: 50.17. Result ID: 69cb8e9347f34ce924b410c9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72f11e209131cea5970b.md>)

Recorded evaluation configuration: {"Model":"Llama3.1-8B","Avg. Bits":"4.25","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

### INT6

ARC-C Accuracy: 48.38. Result ID: 69cb8e9347f34ce924b410cc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

Recorded evaluation configuration: {"Model":"Llama3.1-8B","Avg. Bits":"6.00","Evaluation Protocol":"0-shot","Quantization Implementation":"symmetric per-token"}. This record does not assert independent reproduction.

### QuaRot

ARC-C Accuracy: 46.42. Result ID: 69cb8e9347f34ce924b410c7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa756a1e209131cea67b8c.md>)

[Implementation](<https://github.com/spcl/QuaRot>)

Recorded evaluation configuration: {"Model":"Llama3.1-8B","Avg. Bits":"4.12","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

### QUIK

ARC-C Accuracy: 44.62. Result ID: 69cb8e9347f34ce924b410c8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71141e209131cea4f205.md>)

Recorded evaluation configuration: {"Model":"Llama3.1-8B","Avg. Bits":"5.95","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

### AMXFP4

ARC-C Accuracy: 43.77. Result ID: 69cb8e9347f34ce924b410cb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a6a1e209131cea83140.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78f21e209131cea7b600.md>)

[Implementation](<https://www.sota2.com/aiha-lab/MX-QLLM.git>)

Recorded evaluation configuration: {"Model":"Llama3.1-8B","Avg. Bits":"5.00","Evaluation Protocol":"0-shot"}. This record does not assert independent reproduction.

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