# Tool-selection bias evaluation on Tool-selection bias benchmark

Source: [Tool-selection bias evaluation on Tool-selection bias benchmark](<https://www.sota2.com/research/sota/tool-selection-bias-evaluation-on-tool-selection-bias-benchmark>)

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

Data updated: 2026-03-12T07:03:26.238Z

Evaluation focusing on measuring bias related to tool selection capabilities, quantified by metrics such as δAPI, δpos, and δmodel.

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 | Tool-selection bias evaluation |
| Dataset (including any stated split) | Tool-selection bias benchmark |
| Metric | δAPI Bias |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 7 of 7 |
| 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

- [δAPI Bias](<https://www.sota2.com/research/sota/tool-selection-bias-evaluation-on-tool-selection-bias-benchmark.md?metric=%CE%B4API%20Bias>): Lower is better; unit: SCALAR.

- [δpos Bias](<https://www.sota2.com/research/sota/tool-selection-bias-evaluation-on-tool-selection-bias-benchmark.md?metric=%CE%B4pos%20Bias>): Lower is better; unit: SCALAR.

- [δmodel Bias](<https://www.sota2.com/research/sota/tool-selection-bias-evaluation-on-tool-selection-bias-benchmark.md?metric=%CE%B4model%20Bias>): Lower is better; unit: SCALAR.

## Recorded results

| Method | δAPI Bias | Result date | Report paper |
| --- | --- | --- | --- |
| DeepSeek-V3.2-Exp | 0.249 | Not recorded | 68fa7bc31e209131cea8a405 |
| ToolLLaMA | 0.277 | Not recorded | 68fa7bc31e209131cea8a405 |
| GPT-3.5-turbo | 0.32 | Not recorded | 68fa7bc31e209131cea8a405 |
| Qwen3 235B | 0.33 | Not recorded | 68fa7bc31e209131cea8a405 |
| GPT-4.1 mini | 0.331 | Not recorded | 68fa7bc31e209131cea8a405 |
| Gemini 2.5 Flash | 0.365 | Not recorded | 68fa7bc31e209131cea8a405 |
| Claude 3.5 Sonnet | 0.37 | Not recorded | 68fa7bc31e209131cea8a405 |

### DeepSeek-V3.2-Exp

δAPI Bias: 0.249. Result ID: 69b263da6dbb7075b95c709f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc31e209131cea8a405.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76a91e209131cea6e3eb.md>)

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

### ToolLLaMA

δAPI Bias: 0.277. Result ID: 69b263da6dbb7075b95c70a1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc31e209131cea8a405.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70ce1e209131cea4d478.md>)

[Implementation](<https://github.com/OpenBMB/ToolBench>)

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

### GPT-3.5-turbo

δAPI Bias: 0.32. Result ID: 69b263da6dbb7075b95c70a2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc31e209131cea8a405.md>)

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

### Qwen3 235B

δAPI Bias: 0.33. Result ID: 69b263da6dbb7075b95c70a0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc31e209131cea8a405.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70b91e209131cea4cdc5.md>)

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

### GPT-4.1 mini

δAPI Bias: 0.331. Result ID: 69b263da6dbb7075b95c70a3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc31e209131cea8a405.md>)

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

### Gemini 2.5 Flash

δAPI Bias: 0.365. Result ID: 69b263da6dbb7075b95c709d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc31e209131cea8a405.md>)

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

### Claude 3.5 Sonnet

δAPI Bias: 0.37. Result ID: 69b263da6dbb7075b95c709e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc31e209131cea8a405.md>)

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

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