# Federated Multi-task Fine-tuning on Multi-task Federated Learning Seq, Token, Choice

Source: [Federated Multi-task Fine-tuning on Multi-task Federated Learning Seq, Token, Choice](<https://www.sota2.com/research/sota/federated-multi-task-fine-tuning-on-multi-task-federated-learning-seq-token-choice>)

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

Data updated: 2026-04-16T07:03:06.378Z

Evaluation of Federated Multi-task Fine-tuning performance on the Multi-task Federated Learning Seq, Token, Choice dataset, reporting Max Reward, Average Accuracy, Latency, and Energy consumption.

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 | Federated Multi-task Fine-tuning |
| Dataset (including any stated split) | Multi-task Federated Learning Seq, Token, Choice |
| Metric | Max Reward |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 4 of 4 |
| 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

- [Max Reward](<https://www.sota2.com/research/sota/federated-multi-task-fine-tuning-on-multi-task-federated-learning-seq-token-choice.md?metric=Max%20Reward>): Higher is better; unit: SCALAR.

- [Accuracy](<https://www.sota2.com/research/sota/federated-multi-task-fine-tuning-on-multi-task-federated-learning-seq-token-choice.md?metric=Accuracy>): Higher is better; unit: PERCENT.

- [Latency (s)](<https://www.sota2.com/research/sota/federated-multi-task-fine-tuning-on-multi-task-federated-learning-seq-token-choice.md?metric=Latency%20(s)>): Lower is better; unit: TIME.

- [Energy (J)](<https://www.sota2.com/research/sota/federated-multi-task-fine-tuning-on-multi-task-federated-learning-seq-token-choice.md?metric=Energy%20(J)>): Lower is better; unit: OTHER.

## Recorded results

| Method | Max Reward | Result date | Report paper |
| --- | --- | --- | --- |
| UCB-DUAL | 483.767 | Not recorded | 68fa7a8f1e209131cea83e56 |
| UCB-DUAL | 481.854 | Not recorded | 68fa7a8f1e209131cea83e56 |
| HetLoRA | 462.448 | Not recorded | 68fa7a8f1e209131cea83e56 |
| HomoLoRA | 446.32 | Not recorded | 68fa7a8f1e209131cea83e56 |

### UCB-DUAL

Max Reward: 483.767. Result ID: 69e0881f47f34ce924c943d2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a8f1e209131cea83e56.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a8f1e209131cea83e56.md>)

Recorded evaluation configuration: {"scheduler":"enabled"}. This record does not assert independent reproduction.

### UCB-DUAL

Max Reward: 481.854. Result ID: 69e0881f47f34ce924c943d1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a8f1e209131cea83e56.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a8f1e209131cea83e56.md>)

Recorded evaluation configuration: {"scheduler":"disabled"}. This record does not assert independent reproduction.

### HetLoRA

Max Reward: 462.448. Result ID: 69e0881f47f34ce924c943d0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a8f1e209131cea83e56.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72481e209131cea5586e.md>)

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

### HomoLoRA

Max Reward: 446.32. Result ID: 69e0881f47f34ce924c943cf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a8f1e209131cea83e56.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e761e209131cea40072.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.