# Scaling Law Parameter Estimation on Scaling Law Analysis Corpus

Source: [Scaling Law Parameter Estimation on Scaling Law Analysis Corpus](<https://www.sota2.com/research/sota/scaling-law-parameter-estimation-on-scaling-law-analysis-corpus>)

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

Data updated: 2026-02-26T06:03:41.225Z

Evaluation of model performance on estimating scaling law coefficients (Coeff. a and Coeff. b) derived from the Scaling Law Analysis Corpus.

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 | Scaling Law Parameter Estimation |
| Dataset (including any stated split) | Scaling Law Analysis Corpus |
| Metric | Scaling Law Coeff. a |
| Unit | SCALAR |
| Direction | Higher 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

- [Scaling Law Coeff. a](<https://www.sota2.com/research/sota/scaling-law-parameter-estimation-on-scaling-law-analysis-corpus.md?metric=Scaling%20Law%20Coeff.%20a>): Higher is better; unit: SCALAR.

- [Scaling Law Coeff. b](<https://www.sota2.com/research/sota/scaling-law-parameter-estimation-on-scaling-law-analysis-corpus.md?metric=Scaling%20Law%20Coeff.%20b>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Scaling Law Coeff. a | Result date | Report paper |
| --- | --- | --- | --- |
| Kaplan et al. | 0.73 | Not recorded | 68fa6e5c1e209131cea3f593 |
| Minimum over training curves (Approach 1) | 0.5 | Not recorded | 68fa6e5c1e209131cea3f593 |
| IsoFLOP profiles (Approach 2) | 0.49 | Not recorded | 68fa6e5c1e209131cea3f593 |
| Parametric modelling of the loss (Approach 3) | 0.46 | Not recorded | 68fa6e5c1e209131cea3f593 |

### Kaplan et al.

Scaling Law Coeff. a: 0.73. Result ID: 699016523a6c314849fe0a87.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e5c1e209131cea3f593.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66eb1e209131cea132ee.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Minimum over training curves (Approach 1)

Scaling Law Coeff. a: 0.5. Result ID: 699016523a6c314849fe0a84.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e5c1e209131cea3f593.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e5c1e209131cea3f593.md>)

Recorded evaluation configuration: {"Estimation Technique":"Bootstrapping (80% dataset, 100 samples)"}. This record does not assert independent reproduction.

### IsoFLOP profiles (Approach 2)

Scaling Law Coeff. a: 0.49. Result ID: 699016523a6c314849fe0a85.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e5c1e209131cea3f593.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e5c1e209131cea3f593.md>)

Recorded evaluation configuration: {"Estimation Technique":"Bootstrapping (80% dataset, 100 samples)"}. This record does not assert independent reproduction.

### Parametric modelling of the loss (Approach 3)

Scaling Law Coeff. a: 0.46. Result ID: 699016523a6c314849fe0a86.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e5c1e209131cea3f593.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e5c1e209131cea3f593.md>)

Recorded evaluation configuration: {"Estimation Technique":"Bootstrapping (80% dataset, 100 samples)"}. This record does not assert independent reproduction.

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