# LLM Pre-training on GPT-2 (val)

Source: [LLM Pre-training on GPT-2 (val)](<https://www.sota2.com/research/sota/llm-pre-training-on-gpt-2-val>)

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

Data updated: 2026-04-20T06:53:52.122Z

Evaluation of LLM Pre-training performance on the validation split of the GPT-2 corpus, measured by Loss and Perplexity.

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 | LLM Pre-training |
| Dataset (including any stated split) | GPT-2 (val) |
| Metric | Loss |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 4 of 4 |
| 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

- [Loss](<https://www.sota2.com/research/sota/llm-pre-training-on-gpt-2-val.md?metric=Loss>): Lower is better; unit: SCALAR.

- [Perplexity](<https://www.sota2.com/research/sota/llm-pre-training-on-gpt-2-val.md?metric=Perplexity>): Lower is better; unit: SCALAR.

## Recorded results

| Method | Loss | Result date | Report paper |
| --- | --- | --- | --- |
| StoSignSGDv2 | 3.405 | Not recorded | 69e5b3057d0172140ec7df4b |
| StoSignSGD | 3.418 | Not recorded | 69e5b3057d0172140ec7df4b |
| Lion | 3.46 | Not recorded | 69e5b3057d0172140ec7df4b |
| SignSGD | 3.546 | Not recorded | 69e5b3057d0172140ec7df4b |

### StoSignSGDv2

Loss: 3.405. Result ID: 69e5ccf547f34ce924cb69b0.

[Reporting paper](<https://www.sota2.com/research/paper/69e5b3057d0172140ec7df4b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e5b3057d0172140ec7df4b.md>)

Recorded evaluation configuration: {"Precision":"FP8"}. This record does not assert independent reproduction.

### StoSignSGD

Loss: 3.418. Result ID: 69e5ccf547f34ce924cb69af.

[Reporting paper](<https://www.sota2.com/research/paper/69e5b3057d0172140ec7df4b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e5b3057d0172140ec7df4b.md>)

Recorded evaluation configuration: {"Precision":"FP8"}. This record does not assert independent reproduction.

### Lion

Loss: 3.46. Result ID: 69e5ccf547f34ce924cb69ae.

[Reporting paper](<https://www.sota2.com/research/paper/69e5b3057d0172140ec7df4b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f421e209131cea448e3.md>)

[Implementation](<https://github.com/google/automl/tree/master/lion>)

Recorded evaluation configuration: {"Precision":"FP8"}. This record does not assert independent reproduction.

### SignSGD

Loss: 3.546. Result ID: 69e5ccf547f34ce924cb69ad.

[Reporting paper](<https://www.sota2.com/research/paper/69e5b3057d0172140ec7df4b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa55e31e209131ce9982b8.md>)

Recorded evaluation configuration: {"Precision":"FP8"}. This record does not assert independent reproduction.

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