EMG-FM-Bench
EMG-FM-Bench: A Comprehensive Benchmark for Foundation Model Transfer and Adaptation on Electromyography
Evaluates 9 foundation models on 20 public EMG datasets (>1M segments) for transfer, adaptation, generalization, and task consistency.
How to run
$ npm i -g lettokenburn-cli$ ltb bench info emg-fm-bench
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
10 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| Chronos-2-B | 32.1 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| TimesFM | 30.6 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| Chronos-2-S | 30.2 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| MOMENT-S | 28.5 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| MOMENT-B | 27.9 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| UniTS | 23.5 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| Lag-Llama | 23.3 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| BiLSTM | 17.5 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| BRANT | 16.3 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
| PatchTST | 13.4 | N/A | Independent | medium | Oct 5, 2026 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsAbout
Unifies 20 public EMG datasets with over 1 million segments to evaluate nine pretrained foundation models. It assesses linear probing vs. fine-tuning, pretraining benefits, cross-user generalization, and task consistency.
Scoring uses macro-F1 across classification and decoding tasks. Findings show full fine-tuning significantly alters performance, pretraining benefits are not universal, and cross-user generalization remains challenging despite five-shot adaptation.