MOPD: Multi-Teacher On-Policy Distillation for Capability Integration in LLM Post-Training
Paper β’ 2606.30406 β’ Published β’ 14
AI & ML interests
None defined yet.
Recent Activity
View all activity
AdinaYΒ
submitted a
paper to Daily Papers 5 days ago
AdinaYΒ
submitted a
paper to Daily Papers 20 days ago
AdinaYΒ
submitted a
paper to Daily Papers 21 days ago
Post
136
π₯ Two medical English ASR models are up
Hey, back from a long holiday. While I was out the team kept working on this one and the results are pretty interesting. Medical English ASR, evaluated against the published MultiMed paper.
π©Ί yuriyvnv/parakeet-tdt-0.6b-EN-Medical
π©Ί yuriyvnv/Qwen3-ASR-1.7B-EN-Medical
Both trained on MultiMed (leduckhai/MultiMed) mixed with Common Voice 17 English train and validation. Mixing CV in prevents catastrophic forgetting of general English. Medical-only training without CV cost us 5 absolute WER points on general English.
π Normalized WER on MultiMed-en test, same protocol as the paper:
Parakeet 0.6B zero-shot: 19.22
Parakeet 0.6B fine-tuned: 14.31 (25% relative reduction)
Qwen3-ASR 1.7B zero-shot: 16.41 (although here we had catastrophic forgetting on CV test set)
Qwen3-ASR 1.7B fine-tuned: 16.50
@hf-audio @QwenLM thanks for the toolkits. Big thanks to @leduckhai and the MultiMed authors for the dataset.
#asr #speech #medical #healthcareai #parakeet #qwen #qwen3asr #nemo #medicalasr
Hey, back from a long holiday. While I was out the team kept working on this one and the results are pretty interesting. Medical English ASR, evaluated against the published MultiMed paper.
π©Ί yuriyvnv/parakeet-tdt-0.6b-EN-Medical
π©Ί yuriyvnv/Qwen3-ASR-1.7B-EN-Medical
Both trained on MultiMed (leduckhai/MultiMed) mixed with Common Voice 17 English train and validation. Mixing CV in prevents catastrophic forgetting of general English. Medical-only training without CV cost us 5 absolute WER points on general English.
π Normalized WER on MultiMed-en test, same protocol as the paper:
Parakeet 0.6B zero-shot: 19.22
Parakeet 0.6B fine-tuned: 14.31 (25% relative reduction)
Qwen3-ASR 1.7B zero-shot: 16.41 (although here we had catastrophic forgetting on CV test set)
Qwen3-ASR 1.7B fine-tuned: 16.50
@hf-audio @QwenLM thanks for the toolkits. Big thanks to @leduckhai and the MultiMed authors for the dataset.
#asr #speech #medical #healthcareai #parakeet #qwen #qwen3asr #nemo #medicalasr
Pending Access Request to Join HF Blog Explorers
6
#17 opened 3 months ago
by 
AINovice2005
Post
3010
ππ»ββοΈ Hey there folks ,
Turns out : if we predict π earth we can save a lot of time looking for interesting things and less time looking at things that we expect to see.
Sentinel-2 imagery π°οΈbasically takes a long time to download towards earth. so our "near real time" systems are quite far from that in practical terms.
meanwhile , if we "predict" what we will see , based on what we do see , we can send down much less data in a timely way , and prioritize π‘earth-bound response .
I'm talking about illegal fishing , logging , mining or building in nature reserves , the more of that we predict early the more we're able to stop it on time.
At least that's the concept !
check out the blog : https://huggingface.co/blog/Tonic/save-patagonia-by-predicting-earth
- Collection: https://huggingface.co/collections/NuTonic/earth-observation-with-temporal-and-general-understanding
- Code: https://github.com/Josephrp/Nutonic
- Dataset: NuTonic/sat-vl-sft-training-ready-v1
- Model: NuTonic/lspace
- Training: NuTonic/lspace-trackio
- Evals: NuTonic/Patagonia_Eval
Turns out : if we predict π earth we can save a lot of time looking for interesting things and less time looking at things that we expect to see.
Sentinel-2 imagery π°οΈbasically takes a long time to download towards earth. so our "near real time" systems are quite far from that in practical terms.
meanwhile , if we "predict" what we will see , based on what we do see , we can send down much less data in a timely way , and prioritize π‘earth-bound response .
I'm talking about illegal fishing , logging , mining or building in nature reserves , the more of that we predict early the more we're able to stop it on time.
At least that's the concept !
check out the blog : https://huggingface.co/blog/Tonic/save-patagonia-by-predicting-earth
- Collection: https://huggingface.co/collections/NuTonic/earth-observation-with-temporal-and-general-understanding
- Code: https://github.com/Josephrp/Nutonic
- Dataset: NuTonic/sat-vl-sft-training-ready-v1
- Model: NuTonic/lspace
- Training: NuTonic/lspace-trackio
- Evals: NuTonic/Patagonia_Eval
- 2 replies
Post
4133
Just published: how we built production Sango (Central African Republic) translation without fine-tuning, parallel corpus, or training compute.
The method β vocabulary-augmented prompting with a 581-entry native-speaker-verified lexicon β generalizes to any of the ~2,000 African languages at the same data-poverty level. Recipe, dataset, and code template all included.
π Blog: https://huggingface.co/blog/MEYNG/sangoai
π¦ Dataset: MEYNG/sango-vocabulary
Would especially value feedback from anyone working on other low-resource African languages β Ewondo, Lingala, Wolof next on our roadmap.
The method β vocabulary-augmented prompting with a 581-entry native-speaker-verified lexicon β generalizes to any of the ~2,000 African languages at the same data-poverty level. Recipe, dataset, and code template all included.
π Blog: https://huggingface.co/blog/MEYNG/sangoai
π¦ Dataset: MEYNG/sango-vocabulary
Would especially value feedback from anyone working on other low-resource African languages β Ewondo, Lingala, Wolof next on our roadmap.
- 2 replies
merveΒ
updated a
Space about 2 months ago
catherinearnettΒ
authored 2
papers about 2 months ago
[Support] Community Articles
π₯π 2
106
#5 opened over 2 years ago
by 
victor
[Support] Community Articles
ππ€ 2
106
#5 opened over 2 years ago
by victor
Post
3713
π The WAVe paper is officially out in the Information Sciences Journal.
You saw the PT and NL model releases earlier this year. This is the peer-reviewed paper behind them, with the full method, ablations, and downstream ASR evaluation.
Quick recap: WAVe is a 1B multimodal embedding model that filters synthetic speech at the word level, not the sentence level. On Portuguese ASR it cuts training steps by 34%, improves cross-domain generalization by 50%, and matches WER with 30% less synthetic data.
π¦ Resources
- Paper: https://www.sciencedirect.com/science/article/pii/S0020025526005220
- PT model: yuriyvnv/WAVe-1B-Multimodal-PT
- NL model: yuriyvnv/WAVe-1B-Multimodal-NL
- Collection: https://huggingface.co/collections/yuriyvnv/multi-modal-embeddings-for-synthetic-transcript-filtering
- Code: https://github.com/yuriyvnv/WAVe
If you train ASR on synthetic or back-translated data, would like to see WAVe benchmarked on other languages.
@reach-vb @ylacombe @hf-audio @BramVanroy
#speech #asr #multimodal #syntheticdata #lowresource
You saw the PT and NL model releases earlier this year. This is the peer-reviewed paper behind them, with the full method, ablations, and downstream ASR evaluation.
Quick recap: WAVe is a 1B multimodal embedding model that filters synthetic speech at the word level, not the sentence level. On Portuguese ASR it cuts training steps by 34%, improves cross-domain generalization by 50%, and matches WER with 30% less synthetic data.
π¦ Resources
- Paper: https://www.sciencedirect.com/science/article/pii/S0020025526005220
- PT model: yuriyvnv/WAVe-1B-Multimodal-PT
- NL model: yuriyvnv/WAVe-1B-Multimodal-NL
- Collection: https://huggingface.co/collections/yuriyvnv/multi-modal-embeddings-for-synthetic-transcript-filtering
- Code: https://github.com/yuriyvnv/WAVe
If you train ASR on synthetic or back-translated data, would like to see WAVe benchmarked on other languages.
@reach-vb @ylacombe @hf-audio @BramVanroy
#speech #asr #multimodal #syntheticdata #lowresource
ritishshriraoΒ
authored 2
papers 2 months ago
Future of Agentic Models
π₯ 2
12
#18 opened 2 months ago
by 
MohamedRashad
π© Report: Spam
1
#19 opened 2 months ago
by 
ccocks-deca
Future of Agentic Models
π₯ 2
12
#18 opened 2 months ago
by MohamedRashad
π© Report: Spam
1
#19 opened 2 months ago
by ccocks-deca
Future of Agentic Models
π₯ 2
12
#18 opened 2 months ago
by MohamedRashad