Skill retrieval on ToolBench
97Precision@RShardMemo
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ShardMemoModel=GPT-OSS-120B, R=3, Scope=Tier-C Skills2026.01 | 97 | 1.94 | |
| Embedding similarityModel=GPT-OSS-120B, R=32026.01 | 88 | 1.81 | |
| Static skill libraryModel=GPT-OSS-120B, R=32026.01 | 86 | 1.76 | |
| BM25 retrievalModel=GPT-OSS-120B, R=32026.01 | 85 | 1.72 | |
| RecencyModel=GPT-OSS-120B, R=32026.01 | 79 | 1.45 | |
| RandomModel=GPT-OSS-120B, R=32026.01 | 71 | 1.58 | |
| Trace kNNModel=GPT-OSS-120B, R=32026.01 | 68 | 1.67 |