unsloth/Qwen3.6-35B-A3B-GGUF
Image-Text-to-Text โข 35B โข Updated โข 1.57M โข 800
Allanatrix PRO
Allanatrix
AI & ML interests
ML engineering/Research
Recent Activity
liked a model 5 days ago
unsloth/Qwen3.6-35B-A3B-GGUF liked a dataset 6 days ago
proxima-fusion/constellaration liked a Space 6 days ago
proxima-fusion/constellaration-benchOrganizations
liked a
model 5 days ago
liked a
dataset 6 days ago
liked a
Space 6 days ago
reacted to anakin87's post with โค๏ธ 6 days ago
Post
10371
How LLM training with RL Environments works?
It all starts with ๐ฅ๐ฒ๐ถ๐ป๐ณ๐ผ๐ฟ๐ฐ๐ฒ๐บ๐ฒ๐ป๐ ๐๐ฒ๐ฎ๐ฟ๐ป๐ถ๐ป๐ด ๐๐ถ๐๐ต ๐ฉ๐ฒ๐ฟ๐ถ๐ณ๐ถ๐ฎ๐ฏ๐น๐ฒ ๐ฅ๐ฒ๐๐ฎ๐ฟ๐ฑ๐
- question asked
- model generates reasoning + answer
- answer checked against ground truth
- reward drives RL training
In this setup, the environment is simple: fixed questions and answers, rollout logic, reward(s)
Consider a more complex tic-tac-toe env โโญ
It adds:
- dynamic game generation/handling
- tunable opponent skill
- multi-turn interactions
(envs can also include tools)
---
What happens at training?
We use ๐๐ฟ๐ผ๐๐ฝ ๐ฅ๐ฒ๐น๐ฎ๐๐ถ๐๐ฒ ๐ฃ๐ผ๐น๐ถ๐ฐ๐ ๐ข๐ฝ๐๐ถ๐บ๐ถ๐๐ฎ๐๐ถ๐ผ๐ป with a tic-tac-toe env
No critic model needed, the group is the baseline
Simpler than PPO
1๏ธโฃ Rollout generation: from the same board, model plays N games via sampling
2๏ธโฃ Each game scored with deterministic rewards (win, format, ...)
3๏ธโฃ Mean score computed across the group
4๏ธโฃ Each rollout's advantage = its score minus the group mean
5๏ธโฃ Model updated to favor trajectories above baseline
๐ Repeat
For a deep dive, check out
๐ฑ https://github.com/anakin87/llm-rl-environments-lil-course
a free hands-on course on RL environments for LLMs
It all starts with ๐ฅ๐ฒ๐ถ๐ป๐ณ๐ผ๐ฟ๐ฐ๐ฒ๐บ๐ฒ๐ป๐ ๐๐ฒ๐ฎ๐ฟ๐ป๐ถ๐ป๐ด ๐๐ถ๐๐ต ๐ฉ๐ฒ๐ฟ๐ถ๐ณ๐ถ๐ฎ๐ฏ๐น๐ฒ ๐ฅ๐ฒ๐๐ฎ๐ฟ๐ฑ๐
- question asked
- model generates reasoning + answer
- answer checked against ground truth
- reward drives RL training
In this setup, the environment is simple: fixed questions and answers, rollout logic, reward(s)
Consider a more complex tic-tac-toe env โโญ
It adds:
- dynamic game generation/handling
- tunable opponent skill
- multi-turn interactions
(envs can also include tools)
---
What happens at training?
We use ๐๐ฟ๐ผ๐๐ฝ ๐ฅ๐ฒ๐น๐ฎ๐๐ถ๐๐ฒ ๐ฃ๐ผ๐น๐ถ๐ฐ๐ ๐ข๐ฝ๐๐ถ๐บ๐ถ๐๐ฎ๐๐ถ๐ผ๐ป with a tic-tac-toe env
No critic model needed, the group is the baseline
Simpler than PPO
1๏ธโฃ Rollout generation: from the same board, model plays N games via sampling
2๏ธโฃ Each game scored with deterministic rewards (win, format, ...)
3๏ธโฃ Mean score computed across the group
4๏ธโฃ Each rollout's advantage = its score minus the group mean
5๏ธโฃ Model updated to favor trajectories above baseline
๐ Repeat
For a deep dive, check out
๐ฑ https://github.com/anakin87/llm-rl-environments-lil-course
a free hands-on course on RL environments for LLMs
- 2 replies
updated a
Space 6 days ago
published a
Space 6 days ago
published an article 6 days ago
Article
How I pre-trained a MS/MS model from scratch
liked a
model 6 days ago
Add MassSpecGym evaluation adapter and safetensors runtime loader
#1 opened 7 days ago
by 
Allanatrix
updated a
model 7 days ago
upvoted a paper 9 days ago
published a
model 12 days ago
liked a
model 18 days ago
upvoted a changelog 26 days ago
Hugging Face Changelog
Storage Buckets for Spaces
โข 129
updated a
dataset 27 days ago
published a
dataset 27 days ago
liked a
Space 27 days ago
upvoted a changelog about 1 month ago
Hugging Face Changelog
Introducing hf-mount
โข 223
reacted to MaziyarPanahi's post with ๐ฅ about 1 month ago
Post
2241
We annotated 119K medical images with two frontier VLMs (Qwen 3.5, Kimi K2.5), cross-validated at 93% agreement, and produced 110K training records, all for under $500. Fine-tuning 3 small models (2-3B params) improved all benchmarks: best model reaches +15.0% average exact match.
Everything is open-sourced: datasets, adapters, and code.
https://huggingface.co/blog/OpenMed/synthvision
Everything is open-sourced: datasets, adapters, and code.
https://huggingface.co/blog/OpenMed/synthvision
- 2 replies
liked a
Space about 1 month ago