# Question Answering and Commonsense Reasoning on LM Eval ARCC, ARCE, HellaSwag, PIQA (test)

Source: [Question Answering and Commonsense Reasoning on LM Eval ARCC, ARCE, HellaSwag, PIQA (test)](<https://www.sota2.com/research/sota/question-answering-and-commonsense-reasoning-on-lm-eval-arcc-arce-hellaswag-piqa-test>)

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

Data updated: 2026-02-26T05:23:39.419Z

Evaluation covering multiple commonsense reasoning and question answering tasks (ARCC, ARCE, HellaSwag, PIQA) using the standard test splits.

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 and Commonsense Reasoning |
| Dataset (including any stated split) | LM Eval ARCC, ARCE, HellaSwag, PIQA 0.4.4 standard (test) |
| Metric | ARCC |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 18 of 18 |
| 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

- [ARCC](<https://www.sota2.com/research/sota/question-answering-and-commonsense-reasoning-on-lm-eval-arcc-arce-hellaswag-piqa-test.md?metric=ARCC>): Higher is better; unit: PERCENT.

- [ARCE](<https://www.sota2.com/research/sota/question-answering-and-commonsense-reasoning-on-lm-eval-arcc-arce-hellaswag-piqa-test.md?metric=ARCE>): Higher is better; unit: PERCENT.

- [HellaSwag](<https://www.sota2.com/research/sota/question-answering-and-commonsense-reasoning-on-lm-eval-arcc-arce-hellaswag-piqa-test.md?metric=HellaSwag>): Higher is better; unit: PERCENT.

- [PIQA](<https://www.sota2.com/research/sota/question-answering-and-commonsense-reasoning-on-lm-eval-arcc-arce-hellaswag-piqa-test.md?metric=PIQA>): Higher is better; unit: PERCENT.

## Recorded results

| Method | ARCC | Result date | Report paper |
| --- | --- | --- | --- |
| Meta 'FP8' | 61.6 | Not recorded | 68fa798a1e209131cea7e211 |
| QTIP | 61.5 | Not recorded | 68fa798a1e209131cea7e211 |
| QTIP | 61.3 | Not recorded | 68fa798a1e209131cea7e211 |
| QTIP | 60.7 | Not recorded | 68fa798a1e209131cea7e211 |
| BF16 | 56.7 | Not recorded | 68fa798a1e209131cea7e211 |
| QTIP | 56.3 | Not recorded | 68fa798a1e209131cea7e211 |
| QTIP | 55.1 | Not recorded | 68fa798a1e209131cea7e211 |
| QTIP | 54.4 | Not recorded | 68fa798a1e209131cea7e211 |
| PV-Tuning | 52.7 | Not recorded | 68fa798a1e209131cea7e211 |
| BF16 | 51.6 | Not recorded | 68fa798a1e209131cea7e211 |
| QTIP | 50.7 | Not recorded | 68fa798a1e209131cea7e211 |
| QTIP | 50.4 | Not recorded | 68fa798a1e209131cea7e211 |
| PV-Tuning | 46.2 | Not recorded | 68fa798a1e209131cea7e211 |
| QTIP | 45.1 | Not recorded | 68fa798a1e209131cea7e211 |
| QTIP | 43.5 | Not recorded | 68fa798a1e209131cea7e211 |
| BF16 | 43.3 | Not recorded | 68fa798a1e209131cea7e211 |
| BF16 | 36 | Not recorded | 68fa798a1e209131cea7e211 |
| QTIP | 34.8 | Not recorded | 68fa798a1e209131cea7e211 |

### Meta 'FP8'

ARCC: 61.6. Result ID: 698b531d8d67a41d0a6d1aff.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

Recorded evaluation configuration: {"Bits":"16","Model":"Llama 3.1 405B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### QTIP

ARCC: 61.5. Result ID: 698b531d8d67a41d0a6d1b01.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Implementation](<https://github.com/Cornell-RelaxML/qtip>)

Recorded evaluation configuration: {"Bits":"3","Model":"Llama 3.1 405B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### QTIP

ARCC: 61.3. Result ID: 698b531d8d67a41d0a6d1b00.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Implementation](<https://github.com/Cornell-RelaxML/qtip>)

Recorded evaluation configuration: {"Bits":"4","Model":"Llama 3.1 405B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### QTIP

ARCC: 60.7. Result ID: 698b531d8d67a41d0a6d1b02.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Implementation](<https://github.com/Cornell-RelaxML/qtip>)

Recorded evaluation configuration: {"Bits":"2","Model":"Llama 3.1 405B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### BF16

ARCC: 56.7. Result ID: 698b531d8d67a41d0a6d1b03.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

Recorded evaluation configuration: {"Bits":"16","Model":"Llama 3.1 70B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### QTIP

ARCC: 56.3. Result ID: 698b531d8d67a41d0a6d1b04.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Implementation](<https://github.com/Cornell-RelaxML/qtip>)

Recorded evaluation configuration: {"Bits":"4","Model":"Llama 3.1 70B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### QTIP

ARCC: 55.1. Result ID: 698b531d8d67a41d0a6d1b05.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Implementation](<https://github.com/Cornell-RelaxML/qtip>)

Recorded evaluation configuration: {"Bits":"3","Model":"Llama 3.1 70B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### QTIP

ARCC: 54.4. Result ID: 698b531d8d67a41d0a6d1b06.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Implementation](<https://github.com/Cornell-RelaxML/qtip>)

Recorded evaluation configuration: {"Bits":"2","Model":"Llama 3.1 70B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### PV-Tuning

ARCC: 52.7. Result ID: 698b531d8d67a41d0a6d1b07.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa736c1e209131cea5c6ad.md>)

Recorded evaluation configuration: {"Bits":"2.01","Model":"Llama 3.1 70B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### BF16

ARCC: 51.6. Result ID: 698b531d8d67a41d0a6d1b08.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

Recorded evaluation configuration: {"Bits":"16","Model":"Llama 3.1 8B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### QTIP

ARCC: 50.7. Result ID: 698b531d8d67a41d0a6d1b09.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Implementation](<https://github.com/Cornell-RelaxML/qtip>)

Recorded evaluation configuration: {"Bits":"4","Model":"Llama 3.1 8B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### QTIP

ARCC: 50.4. Result ID: 698b531d8d67a41d0a6d1b0a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Implementation](<https://github.com/Cornell-RelaxML/qtip>)

Recorded evaluation configuration: {"Bits":"3","Model":"Llama 3.1 8B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### PV-Tuning

ARCC: 46.2. Result ID: 698b531d8d67a41d0a6d1b0c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa736c1e209131cea5c6ad.md>)

Recorded evaluation configuration: {"Bits":"2.07","Model":"Llama 3.1 8B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### QTIP

ARCC: 45.1. Result ID: 698b531d8d67a41d0a6d1b0b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Implementation](<https://github.com/Cornell-RelaxML/qtip>)

Recorded evaluation configuration: {"Bits":"2","Model":"Llama 3.1 8B Instruct","Evaluation Protocol":"Zero-shot"}. This record does not assert independent reproduction.

### QTIP

ARCC: 43.5. Result ID: 698c52dd37894ef7d9cd810f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Implementation](<https://github.com/Cornell-RelaxML/qtip>)

Recorded evaluation configuration: {"Backbone Model":"Llama 3.2 3B","Size (GB)":"2.1","Precision":"4-bit"}. This record does not assert independent reproduction.

### BF16

ARCC: 43.3. Result ID: 698c52dd37894ef7d9cd810e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

Recorded evaluation configuration: {"Backbone Model":"Llama 3.2 3B","Size (GB)":"6","Precision":"16-bit"}. This record does not assert independent reproduction.

### BF16

ARCC: 36. Result ID: 698c52dd37894ef7d9cd8110.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

Recorded evaluation configuration: {"Backbone Model":"Llama 3.2 1B","Size (GB)":"2.4","Precision":"16-bit"}. This record does not assert independent reproduction.

### QTIP

ARCC: 34.8. Result ID: 698c52dd37894ef7d9cd8111.

[Reporting paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa798a1e209131cea7e211.md>)

[Implementation](<https://github.com/Cornell-RelaxML/qtip>)

Recorded evaluation configuration: {"Backbone Model":"Llama 3.2 1B","Size (GB)":"0.97","Precision":"4-bit"}. This record does not assert independent reproduction.

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