# Quantitative evaluation of LLM feedback against human gold standards on 50 SOC analysis reports (test)

Source: [Quantitative evaluation of LLM feedback against human gold standards on 50 SOC analysis reports (test)](<https://www.sota2.com/research/sota/quantitative-evaluation-of-llm-feedback-against-human-gold-standards-on-50-soc-analysis-reports-test>)

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

Data updated: 2026-02-26T06:04:36.924Z

Quantitative evaluation of LLM feedback quality by comparing model outputs against human gold standards using correlation metrics (Spearman, Kendall, Pearson) and RMSE on a test set derived from 50 SOC analysis reports.

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 | Quantitative evaluation of LLM feedback against human gold standards |
| Dataset (including any stated split) | 50 SOC analysis reports (test) |
| Metric | Spearman Correlation (ρ) |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 30 of 30 |
| 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

- [Spearman Correlation (ρ)](<https://www.sota2.com/research/sota/quantitative-evaluation-of-llm-feedback-against-human-gold-standards-on-50-soc-analysis-reports-test.md?metric=Spearman%20Correlation%20(%CF%81)>): Higher is better; unit: SCALAR.

- [Kendall Correlation (τ)](<https://www.sota2.com/research/sota/quantitative-evaluation-of-llm-feedback-against-human-gold-standards-on-50-soc-analysis-reports-test.md?metric=Kendall%20Correlation%20(%CF%84)>): Higher is better; unit: SCALAR.

- [Pearson Correlation (r)](<https://www.sota2.com/research/sota/quantitative-evaluation-of-llm-feedback-against-human-gold-standards-on-50-soc-analysis-reports-test.md?metric=Pearson%20Correlation%20(r)>): Higher is better; unit: SCALAR.

- [RMSE](<https://www.sota2.com/research/sota/quantitative-evaluation-of-llm-feedback-against-human-gold-standards-on-50-soc-analysis-reports-test.md?metric=RMSE>): Lower is better; unit: SCALAR.

## Recorded results

| Method | Spearman Correlation (ρ) | Result date | Report paper |
| --- | --- | --- | --- |
| MESSALA | 0.66 | Not recorded | 695de86a53fdd5c361743adc |
| Method 1 | 0.65 | Not recorded | 695de86a53fdd5c361743adc |
| Method 3 | 0.65 | Not recorded | 695de86a53fdd5c361743adc |
| MESSALA | 0.64 | Not recorded | 695de86a53fdd5c361743adc |
| Method 1 | 0.63 | Not recorded | 695de86a53fdd5c361743adc |
| MESSALA | 0.6 | Not recorded | 695de86a53fdd5c361743adc |
| Method 2 | 0.6 | Not recorded | 695de86a53fdd5c361743adc |
| Method 4 | 0.6 | Not recorded | 695de86a53fdd5c361743adc |
| MESSALA | 0.59 | Not recorded | 695de86a53fdd5c361743adc |
| Method 4 | 0.58 | Not recorded | 695de86a53fdd5c361743adc |
| MESSALA | 0.58 | Not recorded | 695de86a53fdd5c361743adc |
| Method 4 | 0.57 | Not recorded | 695de86a53fdd5c361743adc |
| Method 1 | 0.56 | Not recorded | 695de86a53fdd5c361743adc |
| Method 3 | 0.56 | Not recorded | 695de86a53fdd5c361743adc |
| Method 1 | 0.56 | Not recorded | 695de86a53fdd5c361743adc |
| Method 4 | 0.55 | Not recorded | 695de86a53fdd5c361743adc |
| Method 1 | 0.54 | Not recorded | 695de86a53fdd5c361743adc |
| Method 2 | 0.54 | Not recorded | 695de86a53fdd5c361743adc |
| Method 2 | 0.52 | Not recorded | 695de86a53fdd5c361743adc |
| Method 2 | 0.51 | Not recorded | 695de86a53fdd5c361743adc |
| Method 3 | 0.5 | Not recorded | 695de86a53fdd5c361743adc |
| Method 3 | 0.49 | Not recorded | 695de86a53fdd5c361743adc |
| Method 3 | 0.49 | Not recorded | 695de86a53fdd5c361743adc |
| Method 4 | 0.46 | Not recorded | 695de86a53fdd5c361743adc |
| Method 2 | 0.41 | Not recorded | 695de86a53fdd5c361743adc |
| MESSALA | 0.38 | Not recorded | 695de86a53fdd5c361743adc |
| Method 2 | 0.36 | Not recorded | 695de86a53fdd5c361743adc |
| Method 4 | 0.3 | Not recorded | 695de86a53fdd5c361743adc |
| Method 1 | 0.28 | Not recorded | 695de86a53fdd5c361743adc |
| Method 3 | 0.22 | Not recorded | 695de86a53fdd5c361743adc |

### MESSALA

Spearman Correlation (ρ): 0.66. Result ID: 6996f53e5698665ec441ec6b.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gpt-4.1 mini","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### Method 1

Spearman Correlation (ρ): 0.65. Result ID: 6996f53e5698665ec441ec67.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e911e209131cea40b3d.md>)

Recorded evaluation configuration: {"Model":"gpt-4.1 mini","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### Method 3

Spearman Correlation (ρ): 0.65. Result ID: 6996f53e5698665ec441ec69.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gpt-4.1 mini","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### MESSALA

Spearman Correlation (ρ): 0.64. Result ID: 6996f53e5698665ec441ec66.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gpt-4.1","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### Method 1

Spearman Correlation (ρ): 0.63. Result ID: 6996f53e5698665ec441ec62.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e911e209131cea40b3d.md>)

Recorded evaluation configuration: {"Model":"gpt-4.1","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### MESSALA

Spearman Correlation (ρ): 0.6. Result ID: 6996f53e5698665ec441ec61.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gpt-4o","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### Method 2

Spearman Correlation (ρ): 0.6. Result ID: 6996f53e5698665ec441ec63.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f6b1e209131cea4574f.md>)

Recorded evaluation configuration: {"Model":"gpt-4.1","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### Method 4

Spearman Correlation (ρ): 0.6. Result ID: 6996f53e5698665ec441ec65.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gpt-4.1","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### MESSALA

Spearman Correlation (ρ): 0.59. Result ID: 6996f53e5698665ec441ec70.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gpt-oss","Parameter count":"20B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### Method 4

Spearman Correlation (ρ): 0.58. Result ID: 6996f53e5698665ec441ec6a.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gpt-4.1 mini","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### MESSALA

Spearman Correlation (ρ): 0.58. Result ID: 6996f53e5698665ec441ec75.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"qwen3","Parameter count":"14B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### Method 4

Spearman Correlation (ρ): 0.57. Result ID: 6996f53e5698665ec441ec74.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"qwen3","Parameter count":"14B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### Method 1

Spearman Correlation (ρ): 0.56. Result ID: 6996f53e5698665ec441ec6c.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e911e209131cea40b3d.md>)

Recorded evaluation configuration: {"Model":"gpt-oss","Parameter count":"20B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### Method 3

Spearman Correlation (ρ): 0.56. Result ID: 6996f53e5698665ec441ec6e.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gpt-oss","Parameter count":"20B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### Method 1

Spearman Correlation (ρ): 0.56. Result ID: 6996f53e5698665ec441ec71.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e911e209131cea40b3d.md>)

Recorded evaluation configuration: {"Model":"qwen3","Parameter count":"14B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### Method 4

Spearman Correlation (ρ): 0.55. Result ID: 6996f53e5698665ec441ec60.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gpt-4o","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### Method 1

Spearman Correlation (ρ): 0.54. Result ID: 6996f53e5698665ec441ec5d.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e911e209131cea40b3d.md>)

Recorded evaluation configuration: {"Model":"gpt-4o","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### Method 2

Spearman Correlation (ρ): 0.54. Result ID: 6996f53e5698665ec441ec68.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f6b1e209131cea4574f.md>)

Recorded evaluation configuration: {"Model":"gpt-4.1 mini","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### Method 2

Spearman Correlation (ρ): 0.52. Result ID: 6996f53e5698665ec441ec72.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f6b1e209131cea4574f.md>)

Recorded evaluation configuration: {"Model":"qwen3","Parameter count":"14B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### Method 2

Spearman Correlation (ρ): 0.51. Result ID: 6996f53e5698665ec441ec5e.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f6b1e209131cea4574f.md>)

Recorded evaluation configuration: {"Model":"gpt-4o","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### Method 3

Spearman Correlation (ρ): 0.5. Result ID: 6996f53e5698665ec441ec73.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"qwen3","Parameter count":"14B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### Method 3

Spearman Correlation (ρ): 0.49. Result ID: 6996f53e5698665ec441ec5f.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gpt-4o","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### Method 3

Spearman Correlation (ρ): 0.49. Result ID: 6996f53e5698665ec441ec64.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gpt-4.1","Model Category":"Closed-models"}. This record does not assert independent reproduction.

### Method 4

Spearman Correlation (ρ): 0.46. Result ID: 6996f53e5698665ec441ec6f.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gpt-oss","Parameter count":"20B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### Method 2

Spearman Correlation (ρ): 0.41. Result ID: 6996f53e5698665ec441ec6d.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f6b1e209131cea4574f.md>)

Recorded evaluation configuration: {"Model":"gpt-oss","Parameter count":"20B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### MESSALA

Spearman Correlation (ρ): 0.38. Result ID: 6996f53e5698665ec441ec7a.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gemma3","Parameter count":"12B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### Method 2

Spearman Correlation (ρ): 0.36. Result ID: 6996f53e5698665ec441ec77.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f6b1e209131cea4574f.md>)

Recorded evaluation configuration: {"Model":"gemma3","Parameter count":"12B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### Method 4

Spearman Correlation (ρ): 0.3. Result ID: 6996f53e5698665ec441ec79.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gemma3","Parameter count":"12B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### Method 1

Spearman Correlation (ρ): 0.28. Result ID: 6996f53e5698665ec441ec76.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e911e209131cea40b3d.md>)

Recorded evaluation configuration: {"Model":"gemma3","Parameter count":"12B","Model Category":"Open-models"}. This record does not assert independent reproduction.

### Method 3

Spearman Correlation (ρ): 0.22. Result ID: 6996f53e5698665ec441ec78.

[Reporting paper](<https://www.sota2.com/research/paper/695de86a53fdd5c361743adc.md>)

Recorded evaluation configuration: {"Model":"gemma3","Parameter count":"12B","Model Category":"Open-models"}. This record does not assert independent reproduction.

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