# Zero-shot Evaluation on PiQA, LAMBADA, HellaSwag, and ARC (Easy/Challenge)

Source: [Zero-shot Evaluation on PiQA, LAMBADA, HellaSwag, and ARC (Easy/Challenge)](<https://www.sota2.com/research/sota/zero-shot-evaluation-on-piqa-lambada-hellaswag-and-arc-easy-challenge>)

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

Data updated: 2026-03-25T14:08:16.356Z

Zero-shot evaluation performance across a specific suite of reasoning and language understanding tasks: PiQA, LAMBADA, HellaSwag, and the Easy and Challenge splits of ARC.

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 | Zero-shot Evaluation |
| Dataset (including any stated split) | Evaluation Suite (PiQA, LAMBADA, ARC, HellaSwag) zero-shot |
| Metric | PiQA Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 9 of 9 |
| 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

- [PiQA Accuracy](<https://www.sota2.com/research/sota/zero-shot-evaluation-on-piqa-lambada-hellaswag-and-arc-easy-challenge.md?metric=PiQA%20Accuracy>): Higher is better; unit: PERCENT.

- [LAMBADA Accuracy](<https://www.sota2.com/research/sota/zero-shot-evaluation-on-piqa-lambada-hellaswag-and-arc-easy-challenge.md?metric=LAMBADA%20Accuracy>): Higher is better; unit: PERCENT.

- [ARC (Easy) Accuracy](<https://www.sota2.com/research/sota/zero-shot-evaluation-on-piqa-lambada-hellaswag-and-arc-easy-challenge.md?metric=ARC%20(Easy)%20Accuracy>): Higher is better; unit: PERCENT.

- [ARC (Challenge) Accuracy](<https://www.sota2.com/research/sota/zero-shot-evaluation-on-piqa-lambada-hellaswag-and-arc-easy-challenge.md?metric=ARC%20(Challenge)%20Accuracy>): Higher is better; unit: PERCENT.

- [HellaSwag Accuracy](<https://www.sota2.com/research/sota/zero-shot-evaluation-on-piqa-lambada-hellaswag-and-arc-easy-challenge.md?metric=HellaSwag%20Accuracy>): Higher is better; unit: PERCENT.

- [Average Accuracy](<https://www.sota2.com/research/sota/zero-shot-evaluation-on-piqa-lambada-hellaswag-and-arc-easy-challenge.md?metric=Average%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | PiQA Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Dense (full) | 62.47 | Not recorded | 69c36c180636e5bf4afc6507 |
| SFA (k = 16) | 61.73 | Not recorded | 69c36c180636e5bf4afc6507 |
| Dense (full) | 59.79 | Not recorded | 69c36c180636e5bf4afc6507 |
| Dense (d = 64) | 58.43 | Not recorded | 69c36c180636e5bf4afc6507 |
| SFA (k = 8) | 58.02 | Not recorded | 69c36c180636e5bf4afc6507 |
| Dense (full) | 56.34 | Not recorded | 69c36c180636e5bf4afc6507 |
| Dense (d = 32) | 55.17 | Not recorded | 69c36c180636e5bf4afc6507 |
| SFA (k = 8) | 54.92 | Not recorded | 69c36c180636e5bf4afc6507 |
| Dense (d = 32) | 51.3 | Not recorded | 69c36c180636e5bf4afc6507 |

### Dense (full)

PiQA Accuracy: 62.47. Result ID: 69c3ec4b47f34ce924ada9b6.

[Reporting paper](<https://www.sota2.com/research/paper/69c36c180636e5bf4afc6507.md>)

Recorded evaluation configuration: {"Model":"Qwen3-0.6B","Latency (128k context)":"77.65"}. This record does not assert independent reproduction.

### SFA (k = 16)

PiQA Accuracy: 61.73. Result ID: 69c3ec4b47f34ce924ada9b8.

[Reporting paper](<https://www.sota2.com/research/paper/69c36c180636e5bf4afc6507.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c36c180636e5bf4afc6507.md>)

Recorded evaluation configuration: {"Model":"Qwen3-0.6B","Latency (128k context)":"34.20"}. This record does not assert independent reproduction.

### Dense (full)

PiQA Accuracy: 59.79. Result ID: 69c3ec4b47f34ce924ada9b3.

[Reporting paper](<https://www.sota2.com/research/paper/69c36c180636e5bf4afc6507.md>)

Recorded evaluation configuration: {"Model":"GPT2-350M","Latency (128k context)":"46.78"}. This record does not assert independent reproduction.

### Dense (d = 64)

PiQA Accuracy: 58.43. Result ID: 69c3ec4b47f34ce924ada9b7.

[Reporting paper](<https://www.sota2.com/research/paper/69c36c180636e5bf4afc6507.md>)

Recorded evaluation configuration: {"Model":"Qwen3-0.6B","Latency (128k context)":"30.84"}. This record does not assert independent reproduction.

### SFA (k = 8)

PiQA Accuracy: 58.02. Result ID: 69c3ec4b47f34ce924ada9b5.

[Reporting paper](<https://www.sota2.com/research/paper/69c36c180636e5bf4afc6507.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c36c180636e5bf4afc6507.md>)

Recorded evaluation configuration: {"Model":"GPT2-350M","Latency (128k context)":"23.67"}. This record does not assert independent reproduction.

### Dense (full)

PiQA Accuracy: 56.34. Result ID: 69c3ec4b47f34ce924ada9b0.

[Reporting paper](<https://www.sota2.com/research/paper/69c36c180636e5bf4afc6507.md>)

Recorded evaluation configuration: {"Model":"GPT2-124M","Latency (128k context)":"16.86"}. This record does not assert independent reproduction.

### Dense (d = 32)

PiQA Accuracy: 55.17. Result ID: 69c3ec4b47f34ce924ada9b4.

[Reporting paper](<https://www.sota2.com/research/paper/69c36c180636e5bf4afc6507.md>)

Recorded evaluation configuration: {"Model":"GPT2-350M","Latency (128k context)":"20.58"}. This record does not assert independent reproduction.

### SFA (k = 8)

PiQA Accuracy: 54.92. Result ID: 69c3ec4b47f34ce924ada9b2.

[Reporting paper](<https://www.sota2.com/research/paper/69c36c180636e5bf4afc6507.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c36c180636e5bf4afc6507.md>)

Recorded evaluation configuration: {"Model":"GPT2-124M","Latency (128k context)":"9.41"}. This record does not assert independent reproduction.

### Dense (d = 32)

PiQA Accuracy: 51.3. Result ID: 69c3ec4b47f34ce924ada9b1.

[Reporting paper](<https://www.sota2.com/research/paper/69c36c180636e5bf4afc6507.md>)

Recorded evaluation configuration: {"Model":"GPT2-124M","Latency (128k context)":"7.86"}. This record does not assert independent reproduction.

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