# Physical Reasoning on PIQA (PPL, Acc)

Source: [Physical Reasoning on PIQA (PPL, Acc)](<https://www.sota2.com/research/sota/physical-reasoning-on-piqa-ppl-acc>)

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

Data updated: 2026-04-21T09:37:36.508Z

Evaluation of model performance on physical reasoning and problem-solving tasks using the PIQA dataset, reporting Perplexity (PPL) and Accuracy (Acc).

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 | Physical reasoning and problem-solving |
| Dataset (including any stated split) | PIQA |
| Metric | Perplexity |
| Unit | SCALAR |
| Direction | Lower 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

- [Perplexity](<https://www.sota2.com/research/sota/physical-reasoning-on-piqa-ppl-acc.md?metric=Perplexity>): Lower is better; unit: SCALAR.

- [Accuracy](<https://www.sota2.com/research/sota/physical-reasoning-on-piqa-ppl-acc.md?metric=Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Perplexity | Result date | Report paper |
| --- | --- | --- | --- |
| Base | 3.28 | Not recorded | 69e704c27d0172140ec7eef4 |
| Selective Parameter Optimization | 3.38 | Not recorded | 69e704c27d0172140ec7eef4 |
| LoRA(ν + μ) | 3.45 | Not recorded | 69e704c27d0172140ec7eef4 |
| EWCLoRA | 3.47 | Not recorded | 69e704c27d0172140ec7eef4 |
| LoRA(μ) | 3.54 | Not recorded | 69e704c27d0172140ec7eef4 |
| RSLoRA | 3.65 | Not recorded | 69e704c27d0172140ec7eef4 |
| Selective Parameter Optimization | 4.92 | Not recorded | 69e704c27d0172140ec7eef4 |
| Base | 4.94 | Not recorded | 69e704c27d0172140ec7eef4 |
| EWCLoRA | 4.98 | Not recorded | 69e704c27d0172140ec7eef4 |
| LoRA(μ) | 5.43 | Not recorded | 69e704c27d0172140ec7eef4 |
| LoRA(ν + μ) | 5.58 | Not recorded | 69e704c27d0172140ec7eef4 |
| RSLoRA | 5.71 | Not recorded | 69e704c27d0172140ec7eef4 |

### Base

Perplexity: 3.28. Result ID: 69e7455347f34ce924cde0a5.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

Recorded evaluation configuration: {"Backbone":"GPT-J"}. This record does not assert independent reproduction.

### Selective Parameter Optimization

Perplexity: 3.38. Result ID: 69e7455347f34ce924cde0aa.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

Recorded evaluation configuration: {"Backbone":"GPT-J"}. This record does not assert independent reproduction.

### LoRA(ν + μ)

Perplexity: 3.45. Result ID: 69e7455347f34ce924cde0a7.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b121e209131cea2bf59.md>)

[Implementation](<https://github.com/microsoft/LoRA>)

Recorded evaluation configuration: {"Backbone":"GPT-J"}. This record does not assert independent reproduction.

### EWCLoRA

Perplexity: 3.47. Result ID: 69e7455347f34ce924cde0a9.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71021e209131cea4e7fc.md>)

[Implementation](<https://github.com/szxiangjn/world-model-for-language-model>)

Recorded evaluation configuration: {"Backbone":"GPT-J"}. This record does not assert independent reproduction.

### LoRA(μ)

Perplexity: 3.54. Result ID: 69e7455347f34ce924cde0a6.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b121e209131cea2bf59.md>)

[Implementation](<https://github.com/microsoft/LoRA>)

Recorded evaluation configuration: {"Backbone":"GPT-J"}. This record does not assert independent reproduction.

### RSLoRA

Perplexity: 3.65. Result ID: 69e7455347f34ce924cde0a8.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71641e209131cea51208.md>)

Recorded evaluation configuration: {"Backbone":"GPT-J"}. This record does not assert independent reproduction.

### Selective Parameter Optimization

Perplexity: 4.92. Result ID: 69e7455347f34ce924cde0a4.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

Recorded evaluation configuration: {"Backbone":"LLaMA-3"}. This record does not assert independent reproduction.

### Base

Perplexity: 4.94. Result ID: 69e7455347f34ce924cde09f.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

Recorded evaluation configuration: {"Backbone":"LLaMA-3"}. This record does not assert independent reproduction.

### EWCLoRA

Perplexity: 4.98. Result ID: 69e7455347f34ce924cde0a3.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71021e209131cea4e7fc.md>)

[Implementation](<https://github.com/szxiangjn/world-model-for-language-model>)

Recorded evaluation configuration: {"Backbone":"LLaMA-3"}. This record does not assert independent reproduction.

### LoRA(μ)

Perplexity: 5.43. Result ID: 69e7455347f34ce924cde0a0.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b121e209131cea2bf59.md>)

[Implementation](<https://github.com/microsoft/LoRA>)

Recorded evaluation configuration: {"Backbone":"LLaMA-3"}. This record does not assert independent reproduction.

### LoRA(ν + μ)

Perplexity: 5.58. Result ID: 69e7455347f34ce924cde0a1.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b121e209131cea2bf59.md>)

[Implementation](<https://github.com/microsoft/LoRA>)

Recorded evaluation configuration: {"Backbone":"LLaMA-3"}. This record does not assert independent reproduction.

### RSLoRA

Perplexity: 5.71. Result ID: 69e7455347f34ce924cde0a2.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c27d0172140ec7eef4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71641e209131cea51208.md>)

Recorded evaluation configuration: {"Backbone":"LLaMA-3"}. This record does not assert independent reproduction.

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