# Skill-Augmented Agent Performance Evaluation on SkillsBench 86 runnable tasks (frozen-policy transfer)

Source: [Skill-Augmented Agent Performance Evaluation on SkillsBench 86 runnable tasks (frozen-policy transfer)](<https://www.sota2.com/research/sota/skill-augmented-agent-performance-evaluation-on-skillsbench-86-runnable-tasks-frozen-policy-transfer>)

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

Data updated: 2026-06-09T11:04:29.201Z

Evaluation of agent performance using skill augmentation across the 86 runnable tasks defined in the SkillsBench configuration, specifically focusing on frozen-policy transfer scenarios.

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 | Skill-Augmented Agent Performance Evaluation |
| Dataset (including any stated split) | SkillsBench 86 runnable tasks (frozen-policy transfer) |
| Metric | Valid Task Count |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 15 of 15 |
| 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

- [Valid Task Count](<https://www.sota2.com/research/sota/skill-augmented-agent-performance-evaluation-on-skillsbench-86-runnable-tasks-frozen-policy-transfer.md?metric=Valid%20Task%20Count>): Higher is better; unit: SCALAR.

- [Partial Success Rate](<https://www.sota2.com/research/sota/skill-augmented-agent-performance-evaluation-on-skillsbench-86-runnable-tasks-frozen-policy-transfer.md?metric=Partial%20Success%20Rate>): Higher is better; unit: PERCENT.

- [Reward](<https://www.sota2.com/research/sota/skill-augmented-agent-performance-evaluation-on-skillsbench-86-runnable-tasks-frozen-policy-transfer.md?metric=Reward>): Higher is better; unit: SCALAR.

- [QR Score](<https://www.sota2.com/research/sota/skill-augmented-agent-performance-evaluation-on-skillsbench-86-runnable-tasks-frozen-policy-transfer.md?metric=QR%20Score>): Higher is better; unit: SCALAR.

- [RS Score](<https://www.sota2.com/research/sota/skill-augmented-agent-performance-evaluation-on-skillsbench-86-runnable-tasks-frozen-policy-transfer.md?metric=RS%20Score>): Higher is better; unit: SCALAR.

- [RA Score](<https://www.sota2.com/research/sota/skill-augmented-agent-performance-evaluation-on-skillsbench-86-runnable-tasks-frozen-policy-transfer.md?metric=RA%20Score>): Higher is better; unit: SCALAR.

- [Pearson Correlation Coefficient (ρ)](<https://www.sota2.com/research/sota/skill-augmented-agent-performance-evaluation-on-skillsbench-86-runnable-tasks-frozen-policy-transfer.md?metric=Pearson%20Correlation%20Coefficient%20(%CF%81)>): Higher is better; unit: SCALAR.

- [Error Rate (EO)](<https://www.sota2.com/research/sota/skill-augmented-agent-performance-evaluation-on-skillsbench-86-runnable-tasks-frozen-policy-transfer.md?metric=Error%20Rate%20(EO)>): Lower is better; unit: PERCENT.

- [Performance Delta (Δ)](<https://www.sota2.com/research/sota/skill-augmented-agent-performance-evaluation-on-skillsbench-86-runnable-tasks-frozen-policy-transfer.md?metric=Performance%20Delta%20(%CE%94)>): Higher is better; unit: SCALAR.

- [NLD Score](<https://www.sota2.com/research/sota/skill-augmented-agent-performance-evaluation-on-skillsbench-86-runnable-tasks-frozen-policy-transfer.md?metric=NLD%20Score>): Lower is better; unit: PERCENT.

## Recorded results

| Method | Valid Task Count | Result date | Report paper |
| --- | --- | --- | --- |
| Policy-selected | 79 | Not recorded | 6a279e8f90966ade919bfc68 |
| Policy-selected | 78 | Not recorded | 6a279e8f90966ade919bfc68 |
| Rule/formula anchored | 78 | Not recorded | 6a279e8f90966ade919bfc68 |
| Workflow-guarded | 78 | Not recorded | 6a279e8f90966ade919bfc68 |
| Rule/formula anchored | 77 | Not recorded | 6a279e8f90966ade919bfc68 |
| API/code anchored | 77 | Not recorded | 6a279e8f90966ade919bfc68 |
| API/code anchored | 76 | Not recorded | 6a279e8f90966ade919bfc68 |
| Original skills | 76 | Not recorded | 6a279e8f90966ade919bfc68 |
| Policy-selected | 75 | Not recorded | 6a279e8f90966ade919bfc68 |
| Workflow-guarded | 75 | Not recorded | 6a279e8f90966ade919bfc68 |
| Rule/formula anchored | 74 | Not recorded | 6a279e8f90966ade919bfc68 |
| Original skills | 74 | Not recorded | 6a279e8f90966ade919bfc68 |
| API/code anchored | 73 | Not recorded | 6a279e8f90966ade919bfc68 |
| Workflow-guarded | 73 | Not recorded | 6a279e8f90966ade919bfc68 |
| Original skills | 72 | Not recorded | 6a279e8f90966ade919bfc68 |

### Policy-selected

Valid Task Count: 79. Result ID: 6a27f2b1fbc373592c259dc8.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

Recorded evaluation configuration: {"Setting":"Claude Code + Opus 4.6"}. This record does not assert independent reproduction.

### Policy-selected

Valid Task Count: 78. Result ID: 6a27f2b1fbc373592c259dc3.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

Recorded evaluation configuration: {"Setting":"Codex + GPT-5.4"}. This record does not assert independent reproduction.

### Rule/formula anchored

Valid Task Count: 78. Result ID: 6a27f2b1fbc373592c259dc6.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

Recorded evaluation configuration: {"Setting":"Claude Code + Opus 4.6"}. This record does not assert independent reproduction.

### Workflow-guarded

Valid Task Count: 78. Result ID: 6a27f2b1fbc373592c259dc7.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

Recorded evaluation configuration: {"Setting":"Claude Code + Opus 4.6"}. This record does not assert independent reproduction.

### Rule/formula anchored

Valid Task Count: 77. Result ID: 6a27f2b1fbc373592c259dc1.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

Recorded evaluation configuration: {"Setting":"Codex + GPT-5.4"}. This record does not assert independent reproduction.

### API/code anchored

Valid Task Count: 77. Result ID: 6a27f2b1fbc373592c259dc5.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

Recorded evaluation configuration: {"Setting":"Claude Code + Opus 4.6"}. This record does not assert independent reproduction.

### API/code anchored

Valid Task Count: 76. Result ID: 6a27f2b1fbc373592c259dc0.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

Recorded evaluation configuration: {"Setting":"Codex + GPT-5.4"}. This record does not assert independent reproduction.

### Original skills

Valid Task Count: 76. Result ID: 6a27f2b1fbc373592c259dc4.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6992a45f9ebc499c97406aea.md>)

[Implementation](<https://github.com/benchflow-ai/skillsbench>)

Recorded evaluation configuration: {"Setting":"Claude Code + Opus 4.6"}. This record does not assert independent reproduction.

### Policy-selected

Valid Task Count: 75. Result ID: 6a27f2b1fbc373592c259dbe.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

Recorded evaluation configuration: {"Setting":"Gemini-CLI + Gemini 3 Pro"}. This record does not assert independent reproduction.

### Workflow-guarded

Valid Task Count: 75. Result ID: 6a27f2b1fbc373592c259dc2.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

Recorded evaluation configuration: {"Setting":"Codex + GPT-5.4"}. This record does not assert independent reproduction.

### Rule/formula anchored

Valid Task Count: 74. Result ID: 6a27f2b1fbc373592c259dbc.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

Recorded evaluation configuration: {"Setting":"Gemini-CLI + Gemini 3 Pro"}. This record does not assert independent reproduction.

### Original skills

Valid Task Count: 74. Result ID: 6a27f2b1fbc373592c259dbf.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6992a45f9ebc499c97406aea.md>)

[Implementation](<https://github.com/benchflow-ai/skillsbench>)

Recorded evaluation configuration: {"Setting":"Codex + GPT-5.4"}. This record does not assert independent reproduction.

### API/code anchored

Valid Task Count: 73. Result ID: 6a27f2b1fbc373592c259dbb.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

Recorded evaluation configuration: {"Setting":"Gemini-CLI + Gemini 3 Pro"}. This record does not assert independent reproduction.

### Workflow-guarded

Valid Task Count: 73. Result ID: 6a27f2b1fbc373592c259dbd.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

Recorded evaluation configuration: {"Setting":"Gemini-CLI + Gemini 3 Pro"}. This record does not assert independent reproduction.

### Original skills

Valid Task Count: 72. Result ID: 6a27f2b1fbc373592c259dba.

[Reporting paper](<https://www.sota2.com/research/paper/6a279e8f90966ade919bfc68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6992a45f9ebc499c97406aea.md>)

[Implementation](<https://github.com/benchflow-ai/skillsbench>)

Recorded evaluation configuration: {"Setting":"Gemini-CLI + Gemini 3 Pro"}. This record does not assert independent reproduction.

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