# Multi-hop Question Answering on HotpotQA (F1 Score)

Source: [Multi-hop Question Answering on HotpotQA (F1 Score)](<https://www.sota2.com/research/sota/multi-hop-question-answering-on-hotpotqa-f1-score>)

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

Data updated: 2026-05-11T07:32:29.110Z

Evaluation of model performance on the HotpotQA dataset for multi-hop question answering, specifically reporting the F1 Score.

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 | Multi-hop Question Answering |
| Dataset (including any stated split) | HotpotQA |
| Metric | F1 Score |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 25 of 25 |
| 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

- [F1 Score](<https://www.sota2.com/research/sota/multi-hop-question-answering-on-hotpotqa-f1-score.md?metric=F1%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | F1 Score | Result date | Report paper |
| --- | --- | --- | --- |
| SMD | 56.2 | Not recorded | 6a01628bdbc836b87d823083 |
| GRPO Dense | 55.8 | Not recorded | 6a01628bdbc836b87d823083 |
| Sparse-RL | 54.5 | Not recorded | 6a01628bdbc836b87d823083 |
| SMD | 52.6 | Not recorded | 6a01628bdbc836b87d823083 |
| GRPO Dense | 52.4 | Not recorded | 6a01628bdbc836b87d823083 |
| SMD | 51.5 | Not recorded | 6a01628bdbc836b87d823083 |
| Base | 51.2 | Not recorded | 6a01628bdbc836b87d823083 |
| GRPO Dense | 50.6 | Not recorded | 6a01628bdbc836b87d823083 |
| Sparse-RL | 50 | Not recorded | 6a01628bdbc836b87d823083 |
| Sparse-RL | 48.6 | Not recorded | 6a01628bdbc836b87d823083 |
| GRPO w/ SnapKV | 48 | Not recorded | 6a01628bdbc836b87d823083 |
| GRPO w/ SnapKV | 46.5 | Not recorded | 6a01628bdbc836b87d823083 |
| Base | 45.2 | Not recorded | 6a01628bdbc836b87d823083 |
| Base | 45.1 | Not recorded | 6a01628bdbc836b87d823083 |
| SMD | 41.6 | Not recorded | 6a01628bdbc836b87d823083 |
| GRPO w/ SnapKV | 41.5 | Not recorded | 6a01628bdbc836b87d823083 |
| GRPO Dense | 41 | Not recorded | 6a01628bdbc836b87d823083 |
| Sparse-RL | 40.2 | Not recorded | 6a01628bdbc836b87d823083 |
| GRPO w/ SnapKV | 36.5 | Not recorded | 6a01628bdbc836b87d823083 |
| SMD | 36.1 | Not recorded | 6a01628bdbc836b87d823083 |
| GRPO Dense | 35.4 | Not recorded | 6a01628bdbc836b87d823083 |
| Base | 33.5 | Not recorded | 6a01628bdbc836b87d823083 |
| Sparse-RL | 33.2 | Not recorded | 6a01628bdbc836b87d823083 |
| Base | 30.1 | Not recorded | 6a01628bdbc836b87d823083 |
| GRPO w/ SnapKV | 29.8 | Not recorded | 6a01628bdbc836b87d823083 |

### SMD

F1 Score: 56.2. Result ID: 6a0185f747f34ce924e300e5.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7B","Rollout Method":"SMD","Toks. saving":"39.4%"}. This record does not assert independent reproduction.

### GRPO Dense

F1 Score: 55.8. Result ID: 6a0185f747f34ce924e300e2.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73101e209131cea5a395.md>)

[Implementation](<https://github.com/deepseek-ai/DeepSeek-Math>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7B","Rollout Method":"Dense"}. This record does not assert independent reproduction.

### Sparse-RL

F1 Score: 54.5. Result ID: 6a0185f747f34ce924e300e4.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa49541e209131ce9365be.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7B","Rollout Method":"Sparse-RL","Toks. saving":"39.4%"}. This record does not assert independent reproduction.

### SMD

F1 Score: 52.6. Result ID: 6a0185f747f34ce924e300d1.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Qwen3-4B-Instruct","Rollout Method":"SMD","Toks. saving":"50.0%"}. This record does not assert independent reproduction.

### GRPO Dense

F1 Score: 52.4. Result ID: 6a0185f747f34ce924e300ce.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73101e209131cea5a395.md>)

[Implementation](<https://github.com/deepseek-ai/DeepSeek-Math>)

Recorded evaluation configuration: {"Model":"Qwen3-4B-Instruct","Rollout Method":"Dense"}. This record does not assert independent reproduction.

### SMD

F1 Score: 51.5. Result ID: 6a0185f747f34ce924e300e0.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-3B","Rollout Method":"SMD","Toks. saving":"42.0%"}. This record does not assert independent reproduction.

### Base

F1 Score: 51.2. Result ID: 6a0185f747f34ce924e300e1.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7B","Rollout Method":"–"}. This record does not assert independent reproduction.

### GRPO Dense

F1 Score: 50.6. Result ID: 6a0185f747f34ce924e300dd.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73101e209131cea5a395.md>)

[Implementation](<https://github.com/deepseek-ai/DeepSeek-Math>)

Recorded evaluation configuration: {"Model":"Qwen2.5-3B","Rollout Method":"Dense"}. This record does not assert independent reproduction.

### Sparse-RL

F1 Score: 50. Result ID: 6a0185f747f34ce924e300df.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa49541e209131ce9365be.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-3B","Rollout Method":"Sparse-RL","Toks. saving":"42.0%"}. This record does not assert independent reproduction.

### Sparse-RL

F1 Score: 48.6. Result ID: 6a0185f747f34ce924e300d0.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa49541e209131ce9365be.md>)

Recorded evaluation configuration: {"Model":"Qwen3-4B-Instruct","Rollout Method":"Sparse-RL","Toks. saving":"50.0%"}. This record does not assert independent reproduction.

### GRPO w/ SnapKV

F1 Score: 48. Result ID: 6a0185f747f34ce924e300e3.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7B","Rollout Method":"w/ SnapKV"}. This record does not assert independent reproduction.

### GRPO w/ SnapKV

F1 Score: 46.5. Result ID: 6a0185f747f34ce924e300de.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-3B","Rollout Method":"w/ SnapKV"}. This record does not assert independent reproduction.

### Base

F1 Score: 45.2. Result ID: 6a0185f747f34ce924e300cd.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Qwen3-4B-Instruct","Rollout Method":"–"}. This record does not assert independent reproduction.

### Base

F1 Score: 45.1. Result ID: 6a0185f747f34ce924e300dc.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-3B","Rollout Method":"–"}. This record does not assert independent reproduction.

### SMD

F1 Score: 41.6. Result ID: 6a0185f747f34ce924e300db.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-1.5B","Rollout Method":"SMD","Toks. saving":"43.3%"}. This record does not assert independent reproduction.

### GRPO w/ SnapKV

F1 Score: 41.5. Result ID: 6a0185f747f34ce924e300cf.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Qwen3-4B-Instruct","Rollout Method":"w/ SnapKV"}. This record does not assert independent reproduction.

### GRPO Dense

F1 Score: 41. Result ID: 6a0185f747f34ce924e300d8.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73101e209131cea5a395.md>)

[Implementation](<https://github.com/deepseek-ai/DeepSeek-Math>)

Recorded evaluation configuration: {"Model":"Qwen2.5-1.5B","Rollout Method":"Dense"}. This record does not assert independent reproduction.

### Sparse-RL

F1 Score: 40.2. Result ID: 6a0185f747f34ce924e300da.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa49541e209131ce9365be.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-1.5B","Rollout Method":"Sparse-RL","Toks. saving":"43.3%"}. This record does not assert independent reproduction.

### GRPO w/ SnapKV

F1 Score: 36.5. Result ID: 6a0185f747f34ce924e300d9.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-1.5B","Rollout Method":"w/ SnapKV"}. This record does not assert independent reproduction.

### SMD

F1 Score: 36.1. Result ID: 6a0185f747f34ce924e300d6.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Llama-3.2-1B-Instruct","Rollout Method":"SMD","Toks. saving":"45.0%"}. This record does not assert independent reproduction.

### GRPO Dense

F1 Score: 35.4. Result ID: 6a0185f747f34ce924e300d3.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73101e209131cea5a395.md>)

[Implementation](<https://github.com/deepseek-ai/DeepSeek-Math>)

Recorded evaluation configuration: {"Model":"Llama-3.2-1B-Instruct","Rollout Method":"Dense"}. This record does not assert independent reproduction.

### Base

F1 Score: 33.5. Result ID: 6a0185f747f34ce924e300d7.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-1.5B","Rollout Method":"–"}. This record does not assert independent reproduction.

### Sparse-RL

F1 Score: 33.2. Result ID: 6a0185f747f34ce924e300d5.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa49541e209131ce9365be.md>)

Recorded evaluation configuration: {"Model":"Llama-3.2-1B-Instruct","Rollout Method":"Sparse-RL","Toks. saving":"45.0%"}. This record does not assert independent reproduction.

### Base

F1 Score: 30.1. Result ID: 6a0185f747f34ce924e300d2.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Llama-3.2-1B-Instruct","Rollout Method":"–"}. This record does not assert independent reproduction.

### GRPO w/ SnapKV

F1 Score: 29.8. Result ID: 6a0185f747f34ce924e300d4.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628bdbc836b87d823083.md>)

Recorded evaluation configuration: {"Model":"Llama-3.2-1B-Instruct","Rollout Method":"w/ SnapKV"}. This record does not assert independent reproduction.

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