# Tool Retrieval on TOOLRET Zero-Shot Macro-Avg

Source: [Tool Retrieval on TOOLRET Zero-Shot Macro-Avg](<https://www.sota2.com/research/sota/tool-retrieval-on-toolret-zero-shot-macro-avg>)

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

Data updated: 2026-02-26T06:11:09.202Z

Evaluation of model performance on tool retrieval capabilities using the TOOLRET Zero-Shot Macro-Avg setting, measured by nDCG@10 and Completeness@10.

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 Retrieval |
| Dataset (including any stated split) | TOOLRET Zero-Shot Macro-Avg |
| Metric | nDCG@10 |
| Unit | PERCENT |
| 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

- [nDCG@10](<https://www.sota2.com/research/sota/tool-retrieval-on-toolret-zero-shot-macro-avg.md?metric=nDCG%4010>): Higher is better; unit: PERCENT.

- [Completeness@10](<https://www.sota2.com/research/sota/tool-retrieval-on-toolret-zero-shot-macro-avg.md?metric=Completeness%4010>): Higher is better; unit: PERCENT.

## Recorded results

| Method | nDCG@10 | Result date | Report paper |
| --- | --- | --- | --- |
| TOOLQP | 36.9 | Not recorded | 6965d1b66343fa78c8323967 |
| bge-gemma | 34 | Not recorded | 6965d1b66343fa78c8323967 |
| Q2P/SFT | 33.2 | Not recorded | 6965d1b66343fa78c8323967 |
| Q2E/ZS | 32 | Not recorded | 6965d1b66343fa78c8323967 |
| gte-Qwen/ContrastiveFT | 31.5 | Not recorded | 6965d1b66343fa78c8323967 |
| Q2D/PRF | 31.2 | Not recorded | 6965d1b66343fa78c8323967 |
| Q2E/PRF | 31 | Not recorded | 6965d1b66343fa78c8323967 |
| Re-Invoke | 30.3 | Not recorded | 6965d1b66343fa78c8323967 |
| bge-m3 | 30.1 | Not recorded | 6965d1b66343fa78c8323967 |
| Base Retriever (gte-Qwen) | 29.8 | Not recorded | 6965d1b66343fa78c8323967 |
| D2Q | 28.4 | Not recorded | 6965d1b66343fa78c8323967 |
| HyDE/PRF | 27.6 | Not recorded | 6965d1b66343fa78c8323967 |
| Q2D/ZS | 26.7 | Not recorded | 6965d1b66343fa78c8323967 |
| TOOLQP-FORMAT | 26.1 | Not recorded | 6965d1b66343fa78c8323967 |
| HyDE/ZS | 19.9 | Not recorded | 6965d1b66343fa78c8323967 |

### TOOLQP

nDCG@10: 36.9. Result ID: 69968b8c5698665ec4d43854.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

Recorded evaluation configuration: {"Method Category":"Fine-tuning","Training Data Size":"10k","Backbone Model":"1.7B"}. This record does not assert independent reproduction.

### bge-gemma

nDCG@10: 34. Result ID: 69968b8c5698665ec4d43850.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Implementation](<https://github.com/FlagOpen/FlagEmbedding>)

Recorded evaluation configuration: {"Method Category":"Re-ranking with cross-encoder"}. This record does not assert independent reproduction.

### Q2P/SFT

nDCG@10: 33.2. Result ID: 69968b8c5698665ec4d43851.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

Recorded evaluation configuration: {"Method Category":"Fine-tuning","Training Data Size":"10k","Backbone Model":"1.5B/1.7B models"}. This record does not assert independent reproduction.

### Q2E/ZS

nDCG@10: 32. Result ID: 69968b8c5698665ec4d43847.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f491e209131cea44ca0.md>)

Recorded evaluation configuration: {"Method Category":"Prompting","Backbone Model":"Qwen3-30B-A3B-Instruct"}. This record does not assert independent reproduction.

### gte-Qwen/ContrastiveFT

nDCG@10: 31.5. Result ID: 69968b8c5698665ec4d43852.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

Recorded evaluation configuration: {"Method Category":"Fine-tuning","Training Data Size":"10k","Backbone Model":"1.5B/1.7B models"}. This record does not assert independent reproduction.

### Q2D/PRF

nDCG@10: 31.2. Result ID: 69968b8c5698665ec4d4384d.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70d71e209131cea4d93f.md>)

Recorded evaluation configuration: {"Method Category":"Prompting","Backbone Model":"Qwen3-30B-A3B-Instruct"}. This record does not assert independent reproduction.

### Q2E/PRF

nDCG@10: 31. Result ID: 69968b8c5698665ec4d4384c.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f491e209131cea44ca0.md>)

Recorded evaluation configuration: {"Method Category":"Prompting","Backbone Model":"Qwen3-30B-A3B-Instruct"}. This record does not assert independent reproduction.

### Re-Invoke

nDCG@10: 30.3. Result ID: 69968b8c5698665ec4d4384b.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74e81e209131cea649ff.md>)

Recorded evaluation configuration: {"Method Category":"Prompting","Backbone Model":"Qwen3-30B-A3B-Instruct"}. This record does not assert independent reproduction.

### bge-m3

nDCG@10: 30.1. Result ID: 69968b8c5698665ec4d4384f.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Implementation](<https://github.com/FlagOpen/FlagEmbedding>)

Recorded evaluation configuration: {"Method Category":"Re-ranking with cross-encoder"}. This record does not assert independent reproduction.

### Base Retriever (gte-Qwen)

nDCG@10: 29.8. Result ID: 69968b8c5698665ec4d43846.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa703b1e209131cea49c20.md>)

Recorded evaluation configuration: {"Method Category":"Base Retriever"}. This record does not assert independent reproduction.

### D2Q

nDCG@10: 28.4. Result ID: 69968b8c5698665ec4d4384a.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66791e209131cea10690.md>)

[Implementation](<https://github.com/nyu-dl/dl4ir-doc2query>)

Recorded evaluation configuration: {"Method Category":"Prompting","Backbone Model":"Qwen3-30B-A3B-Instruct"}. This record does not assert independent reproduction.

### HyDE/PRF

nDCG@10: 27.6. Result ID: 69968b8c5698665ec4d4384e.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e3b1e209131cea3e835.md>)

[Implementation](<https://github.com/texttron/hyde>)

Recorded evaluation configuration: {"Method Category":"Prompting","Backbone Model":"Qwen3-30B-A3B-Instruct"}. This record does not assert independent reproduction.

### Q2D/ZS

nDCG@10: 26.7. Result ID: 69968b8c5698665ec4d43848.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70d71e209131cea4d93f.md>)

Recorded evaluation configuration: {"Method Category":"Prompting","Backbone Model":"Qwen3-30B-A3B-Instruct"}. This record does not assert independent reproduction.

### TOOLQP-FORMAT

nDCG@10: 26.1. Result ID: 69968b8c5698665ec4d43853.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

Recorded evaluation configuration: {"Method Category":"Fine-tuning","Training Data Size":"10k","Backbone Model":"1.7B"}. This record does not assert independent reproduction.

### HyDE/ZS

nDCG@10: 19.9. Result ID: 69968b8c5698665ec4d43849.

[Reporting paper](<https://www.sota2.com/research/paper/6965d1b66343fa78c8323967.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e3b1e209131cea3e835.md>)

[Implementation](<https://github.com/texttron/hyde>)

Recorded evaluation configuration: {"Method Category":"Prompting","Backbone Model":"Qwen3-30B-A3B-Instruct"}. This record does not assert independent reproduction.

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