# Computational Cost (GFLOPs/token) on Natural Language Generation for XSum

Source: [Computational Cost (GFLOPs/token) on Natural Language Generation for XSum](<https://www.sota2.com/research/sota/computational-cost-gflops-token-on-natural-language-generation-for-xsum>)

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

Data updated: 2026-02-26T05:45:43.176Z

Evaluation of the computational efficiency, specifically GFLOPs per token, required for Natural Language Generation models tested on the XSum dataset.

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 | Natural Language Generation |
| Dataset (including any stated split) | XSum |
| Metric | GFLOPs/token |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 5 of 5 |
| 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

- [GFLOPs/token](<https://www.sota2.com/research/sota/computational-cost-gflops-token-on-natural-language-generation-for-xsum.md?metric=GFLOPs%2Ftoken>): Lower is better; unit: SCALAR.

## Recorded results

| Method | GFLOPs/token | Result date | Report paper |
| --- | --- | --- | --- |
| LTE | 5.38 | Not recorded | 68fa76221e209131cea6b680 |
| Deja Vu | 7.92 | Not recorded | 68fa76221e209131cea6b680 |
| R-LLaMA+MoE | 8.27 | Not recorded | 68fa76221e209131cea6b680 |
| MoEfication | 10.45 | Not recorded | 68fa76221e209131cea6b680 |
| Full | 12.06 | Not recorded | 68fa76221e209131cea6b680 |

### LTE

GFLOPs/token: 5.38. Result ID: 699425f03a6c3148491b648f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76221e209131cea6b680.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76221e209131cea6b680.md>)

[Implementation](<https://github.com/haizhongzheng/LTE>)

Recorded evaluation configuration: {"Backbone":"LLaMA-7B","Permitted Quality Drop":"ROUGE-L \<= 0.05"}. This record does not assert independent reproduction.

### Deja Vu

GFLOPs/token: 7.92. Result ID: 699425f03a6c3148491b648c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76221e209131cea6b680.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71081e209131cea4ec02.md>)

Recorded evaluation configuration: {"Backbone":"LLaMA-7B","Permitted Quality Drop":"ROUGE-L \<= 0.05"}. This record does not assert independent reproduction.

### R-LLaMA+MoE

GFLOPs/token: 8.27. Result ID: 699425f03a6c3148491b648e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76221e209131cea6b680.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c2a1e209131cea32765.md>)

Recorded evaluation configuration: {"Backbone":"LLaMA-7B","Permitted Quality Drop":"ROUGE-L \<= 0.05"}. This record does not assert independent reproduction.

### MoEfication

GFLOPs/token: 10.45. Result ID: 699425f03a6c3148491b648d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76221e209131cea6b680.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c2a1e209131cea32765.md>)

Recorded evaluation configuration: {"Backbone":"LLaMA-7B","Permitted Quality Drop":"ROUGE-L \<= 0.05"}. This record does not assert independent reproduction.

### Full

GFLOPs/token: 12.06. Result ID: 699425f03a6c3148491b648b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76221e209131cea6b680.md>)

Recorded evaluation configuration: {"Backbone":"LLaMA-7B","Permitted Quality Drop":"ROUGE-L \<= 0.05"}. This record does not assert independent reproduction.

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