# Writing on LongBench-Write (LB)

Source: [Writing on LongBench-Write (LB)](<https://www.sota2.com/research/sota/writing-on-longbench-write-lb>)

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

Data updated: 2026-04-14T12:11:55.534Z

Evaluation of general writing capabilities using the LongBench-Write dataset, measured by the aggregate LB 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 | Writing |
| Dataset (including any stated split) | LongBench-Write (LB) |
| Metric | LB Score |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 11 of 11 |
| 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

- [LB Score](<https://www.sota2.com/research/sota/writing-on-longbench-write-lb.md?metric=LB%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | LB Score | Result date | Report paper |
| --- | --- | --- | --- |
| Claude 3.7 | 97.8 | Not recorded | 69ddca787d0172140ec79f31 |
| Qwen3-32B | 93.2 | Not recorded | 69ddca787d0172140ec79f31 |
| Claude 3.5 | 89.3 | Not recorded | 69ddca787d0172140ec79f31 |
| Qwen2.5-7B-Instruct + SFT + RL-TCER | 86.3 | Not recorded | 69ddca787d0172140ec79f31 |
| Qwen3-8B | 85.2 | Not recorded | 69ddca787d0172140ec79f31 |
| Qwen2.5-7B-Instruct + SFT + RL-EndoR | 84.5 | Not recorded | 69ddca787d0172140ec79f31 |
| GPT-4o | 83.1 | Not recorded | 69ddca787d0172140ec79f31 |
| Qwen2.5-7B-Instruct + SFT | 83.1 | Not recorded | 69ddca787d0172140ec79f31 |
| Qwen2.5-32B-Instruct | 78.8 | Not recorded | 69ddca787d0172140ec79f31 |
| Qwen2.5-7B-Instruct | 74.7 | Not recorded | 69ddca787d0172140ec79f31 |
| Llama3.1-8B-Instruct | 60.3 | Not recorded | 69ddca787d0172140ec79f31 |

### Claude 3.7

LB Score: 97.8. Result ID: 69de2eec47f34ce924c6eb09.

[Reporting paper](<https://www.sota2.com/research/paper/69ddca787d0172140ec79f31.md>)

Recorded evaluation configuration: {"Model Category":"General Closed-source Models"}. This record does not assert independent reproduction.

### Qwen3-32B

LB Score: 93.2. Result ID: 69de2eec47f34ce924c6eb0d.

[Reporting paper](<https://www.sota2.com/research/paper/69ddca787d0172140ec79f31.md>)

Recorded evaluation configuration: {"Model Category":"General Open-source Models"}. This record does not assert independent reproduction.

### Claude 3.5

LB Score: 89.3. Result ID: 69de2eec47f34ce924c6eb08.

[Reporting paper](<https://www.sota2.com/research/paper/69ddca787d0172140ec79f31.md>)

Recorded evaluation configuration: {"Model Category":"General Closed-source Models"}. This record does not assert independent reproduction.

### Qwen2.5-7B-Instruct + SFT + RL-TCER

LB Score: 86.3. Result ID: 69de2eec47f34ce924c6eb11.

[Reporting paper](<https://www.sota2.com/research/paper/69ddca787d0172140ec79f31.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69ddca787d0172140ec79f31.md>)

Recorded evaluation configuration: {"Model Category":"Finetuned Open-source Models","Training Stage":"SFT + RL-TCER"}. This record does not assert independent reproduction.

### Qwen3-8B

LB Score: 85.2. Result ID: 69de2eec47f34ce924c6eb0c.

[Reporting paper](<https://www.sota2.com/research/paper/69ddca787d0172140ec79f31.md>)

Recorded evaluation configuration: {"Model Category":"General Open-source Models"}. This record does not assert independent reproduction.

### Qwen2.5-7B-Instruct + SFT + RL-EndoR

LB Score: 84.5. Result ID: 69de2eec47f34ce924c6eb10.

[Reporting paper](<https://www.sota2.com/research/paper/69ddca787d0172140ec79f31.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b01e209131cea7ee8d.md>)

Recorded evaluation configuration: {"Model Category":"Finetuned Open-source Models","Training Stage":"SFT + RL-EndoR"}. This record does not assert independent reproduction.

### GPT-4o

LB Score: 83.1. Result ID: 69de2eec47f34ce924c6eb07.

[Reporting paper](<https://www.sota2.com/research/paper/69ddca787d0172140ec79f31.md>)

Recorded evaluation configuration: {"Model Category":"General Closed-source Models"}. This record does not assert independent reproduction.

### Qwen2.5-7B-Instruct + SFT

LB Score: 83.1. Result ID: 69de2eec47f34ce924c6eb0f.

[Reporting paper](<https://www.sota2.com/research/paper/69ddca787d0172140ec79f31.md>)

Recorded evaluation configuration: {"Model Category":"Finetuned Open-source Models","Training Stage":"SFT"}. This record does not assert independent reproduction.

### Qwen2.5-32B-Instruct

LB Score: 78.8. Result ID: 69de2eec47f34ce924c6eb0b.

[Reporting paper](<https://www.sota2.com/research/paper/69ddca787d0172140ec79f31.md>)

Recorded evaluation configuration: {"Model Category":"General Open-source Models"}. This record does not assert independent reproduction.

### Qwen2.5-7B-Instruct

LB Score: 74.7. Result ID: 69de2eec47f34ce924c6eb0e.

[Reporting paper](<https://www.sota2.com/research/paper/69ddca787d0172140ec79f31.md>)

Recorded evaluation configuration: {"Model Category":"Finetuned Open-source Models","Training Stage":"Base"}. This record does not assert independent reproduction.

### Llama3.1-8B-Instruct

LB Score: 60.3. Result ID: 69de2eec47f34ce924c6eb0a.

[Reporting paper](<https://www.sota2.com/research/paper/69ddca787d0172140ec79f31.md>)

Recorded evaluation configuration: {"Model Category":"General Open-source Models"}. This record does not assert independent reproduction.

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