Hugging Face's logo
Activity Feed Request to join this org

AI & ML interests

None defined yet.

Recent Activity
qgallouedec  updated a dataset 5 days ago
trl-lib/documentation-images
qgallouedec  updated a model 16 days ago
trl-lib/fused-linear-ce
qgallouedec  published a model 16 days ago
trl-lib/fused-linear-ce
View all activity
sergiopaniego 
posted an update 3 days ago
view post
Post
1848
Can you do RL over taste?

I've spent some time reproducing, in the open, Surya N's idea of training a model to paint with code. It's a coding model that learns to paint watercolours by writing JS code, trained with GRPO. I used TRL and OpenEnv for this, with the whole pipeline running on Hugging Face.

The interesting part is that the reward has no correct answer, unlike a math problem. In this case it's based on the artistic preferences of the person who builds the dataset.

Everything is published: the environment, the reference pool, the trained adapters, every painting of every run with the code that made it, and a write-up with all the decisions, including the ones that went wrong.

Blog post: https://huggingface.co/blog/train-to-paint-with-code
qgallouedec 
updated a dataset 5 days ago
sergiopaniego 
posted an update 9 days ago
view post
Post
390
catching up on some bookmarked reads from the summer, reading Antidoom from @liquidai

small reasoning models get stuck more easily when the task involves a long thinking trace and a hard problem. It starts repeating the same word over and over again ("Wait", "Alternatively"…), each repetition makes the next one likelier, and the generation is spent before it reaches an answer

they measured it, 10.2% of completions for an early LFM2.5-2.6B checkpoint and 22.9% for Qwen3.5-4B at greedy. After training those drop to 1.4% and 1.0%

the fix is FTPO (final token preference optimization). What I like is how narrow it is, it only touches the single token where the loop starts

three ways it differs from DPO:
> trains one token position, mid-generation, instead of whole sequences
> spreads probability across ~20 plausible alternatives instead of swapping one overtrained token for another
> keeps the regularizer in logit space, no softmax, so the rest of the vocabulary stays put

the third one is what makes it usable. If you want to edit one position without disturbing the model, you can't have a loss that reshuffles the other 150k logits on the way

and their explanation abt the result: the training teaches the model nothing new about math or code, it clears the failure mode that was blocking answers the model could already produce

full blog > https://www.liquid.ai/blog/antidoom

FTPO itself comes from Antislop, where it was built to strip overused phrasing. LiquidAI retargeted it to doom loops

and under the hood it's a subclass of TRL's DPOTrainer with compute_loss overridden, around 90 lines of loss and no new trainer

we documented that pattern in TRL's docs
https://huggingface.co/docs/trl/main/en/customization#change-the-training-objective
  • 1 reply
·
qgallouedec 
updated a model 16 days ago
qgallouedec 
published a model 16 days ago
sergiopaniego 
posted an update 17 days ago
view post
Post
264
super interesting new paper from Microsoft "Agent Lightning v1.0: Towards Harnessed Agentic RL" by Zhiyuan He et al.

same idea we've seen already several times: you train the agent inside the real harness it ships with, instead of a reimplementation of it

now that recipe has a name → harnessed agentic RL

paper: huggingface.co/papers/2608.17528

the tricky bit they nail down: one rollout is not one training sample

the harness calls the model many times, so a single episode → a variable number of (prompt, response) rows

you don't even know the batch size until the episode finishes running

its real contribution is being first to systematically map the four problems that fall out of that:

> retokenization + sample merging
> advantage calculation over a variable sample count
> loss normalization at the rollout level, not per sample
> backend scheduling when the batch size is dynamic

and it actually works → plain RL inside the real harness, no reimplementation

Qwen3.5-9B on SWE-bench Verified 41.8 → 56.4 (+14.6), with only ~6k examples

the whole thing is ~3,500 lines, any harness, self-hosted k8s

from our side, we've shared some materials on the same line you may want to check out :)

> Agentic RL: Token-In, Token-Out Done Right: https://huggingface.co/blog/huggingface/tito
> a full worked example, opencode owning its loop trained with GRPO: https://huggingface.co/blog/sergiopaniego/trl-openenv-harness-training
> Harness, Scaffold, and the AI Agent Terms Worth Getting Right: https://huggingface.co/blog/agent-glossary

on a similar line:

https://x.com/SergioPaniego/status/2062911580564496576
qgallouedec 
updated a dataset 25 days ago
sergiopaniego 
posted an update 27 days ago
view post
Post
721
Something I really like when I study a subject is understanding its history, how it reached the point where it is today

I did that exercise for RL in post-training: from RLHF and PPO, to verifiable rewards, to the GRPO family of variants, to agents acting in environments. Everything is backed by what the labs themselves say in their public reports (DeepSeek, Qwen, Kimi, GLM-5, Nemotron, Mistral and more), in their own words

This is the companion piece to Class 3 of our Training Agents series with @burtenshaw . The class explains how GRPO works, with three hands-on experiments. The article shows where the same ideas appear at frontier scale

https://huggingface.co/blog/sergiopaniego/agentic-rl-2026
  • 1 reply
·
sergiopaniego 
posted an update about 1 month ago
view post
Post
312
we just released a new blog "Training a coding agent using the OpenCode harness in remote HF sandboxes with TRL and OpenEnv"

you can take a real coding agent (OpenCode), let it run its own tool loop against real coding problems, and train it with RL on the exact tokens it produced

and every rollout runs in its own remote HF sandbox, so rollouts scale out beyond one machine

the loop:
- OpenCode owns its tool loop inside an OpenEnv sandbox
- an in-sandbox proxy records the real token ids + logprobs, per turn
- a hidden-test verifier scores the result, and that is the reward
- TRL trains with AsyncGRPO, weights sync back to vLLM over NCCL

blog + runnable example: https://huggingface.co/blog/sergiopaniego/trl-openenv-harness-training
  • 3 replies
·
sergiopaniego 
posted an update about 1 month ago
view post
Post
2640
LFM2.5-2.6B just dropped!

and the @liquidai blog comes with some nice details about the training procedure, so let's analyze it.

basically, a full agent training pipeline but compressed into 2.6B

base model → SFT → specialized teachers per domain (SFT + RLVR) → on-policy distillation back into one student → agentic RL

the two most interesting stages

→ MOPD: the student generates, each prompt routes to its domain teacher for token-level feedback. teachers branch from the same SFT checkpoint, so their signal stays close to the student's distribution

→ agentic RL: multi-turn GRPO inside real harnesses (OpenClaw, Hermes Agent), one sandbox per rollout, a proxy captures token-level trajectories while the harness stays a black box

this makes a 2.6B that beats much larger models on instruction following and tool use

SFT, distillation, RL, RL envs: exactly what we're covering in our Training Agents livestream series (next one coming soon!)

→ model: LiquidAI/LFM2.5-2.6B
→ blog: https://www.liquid.ai/blog/lfm2-5-2-6b
→ live series: https://www.youtube.com/playlist?list=PLo2EIpI_JMQvQZm-kVlz4wY1vWF0LBcf5
  • 3 replies
·
sergiopaniego 
posted an update about 1 month ago
view post
Post
2666
Simon Willison (@simonw ) has asked every new model to draw a pelican riding a bicycle for some time now

you look at the drawing and you know. but there is no number, so nothing can train against it, no?

I turned this idea into an rl env in OpenEnv. now, you can eval any model against it, and train against it with TRL

read the details!🤓

https://huggingface.co/blog/sergiopaniego/pelican-env-openenv
  • 2 replies
·
sergiopaniego 
posted an update about 1 month ago
julien-c 
posted an update about 1 month ago
view post
Post
5136
who's working on an NVFP4 version of Kimi-K3?
  • 4 replies
·
sergiopaniego 
posted an update about 1 month ago
view post
Post
2925
quick reminder! 🚨

tomorrow (Tuesday, July 28), we're back with Class 3 of the Training Agents live series

🧠 what: reinforcement learning for training agents (GRPO): how it works, how to implement it in TRL, and end-to-end examples
🗓️ when: Tuesday, July 28 - 🕔 5:00 PM CEST / 8:30 PM IST
📍 where: Live on @huggingface 's X, YouTube, and LinkedIn

live: https://www.youtube.com/watch?v=ztdTed5egrM

class 1: https://x.com/SergioPaniego/status/2069382207618379813
class 2: https://x.com/SergioPaniego/status/2075180665184686187
  • 1 reply
·
sergiopaniego 
posted an update about 1 month ago
view post
Post
245
you can now train your own coding agents with trl + openenv, starting with opencode

we just added end-to-end support for training agent harnesses:

> TRL: a loop-owning training path (AsyncGRPOTrainer + HarnessRolloutWorker) that launches the agent in an OpenEnv session, reads back its trace, reconstructs the training samples, and trains with AsyncGRPO
> OpenEnv: the OpenCode harness environment plus a transparent proxy that forwards the agent's model calls and records each turn's token ids and logprobs

you train the actual opencode agent as is, it runs its own loop and tools and the policy learns from the exact tokens it produced

we're shipping a self-contained example: local subprocess sandbox, DeepCoder problems, validated on Qwen3-8B.

> example: https://github.com/huggingface/trl/blob/main/examples/scripts/openenv/opencode.py
> docs: https://huggingface.co/docs/trl/main/openenv

and we're working actively on both sides so expect more 🤓
  • 1 reply
·
sergiopaniego 
posted an update about 2 months ago
sergiopaniego 
posted an update about 2 months ago
view post
Post
263
join us next Tuesday, July 28, for Class 3 of the Training Agents live series!

we'll dive into reinforcement learning for agent training, covering the intuition behind GRPO, how it works, and how to implement it in TRL with practical, e2e examples

see you there 🤠

live: https://www.youtube.com/live/ztdTed5egrM

> in case you missed class 1:
https://x.com/SergioPaniego/status/2069382207618379813
> and in case you missed class 2: https://x.com/SergioPaniego/status/2075180665184686187
sergiopaniego 
posted an update 2 months ago
view post
Post
7794
Frontier models use distillation as a step of their post-training pipelines.

In 2026 it has three jobs: compress a big model into a small one, merge RL experts into a single model, and let a model teach itself.

I wrote up which frontier models use each one and how: https://huggingface.co/blog/sergiopaniego/distillation-2026

It pairs with Class 2 of the Training an Agent series Ben and I are doing, where we teach these techniques hands-on with TRL!
  • 3 replies
·
albertvillanova 
posted an update 2 months ago
view post
Post
3983
🎉 KTO is now part of the stable TRL API

As of Promote KTO to stable API, KTOTrainer and KTOConfig have graduated from trl.experimental to the stable trl API. https://github.com/huggingface/trl/pull/6175

This one closes out a long road. Over the past 6+ months, the "Align KTO with DPO" effort landed ~90 PRs methodically bringing KTO up to the standard we hold for stable trainers, one carefully-scoped change at a time:
- Feature parity with DPO: full VLM support (incl. multi-image), sync_ref_model, PEFT + Liger, ZeRO-3 + PEFT dtype fix, pad_to_multiple_of, activation offloading, IterableDataset and dict eval_dataset, remove_unused_columns, and reference-logprob precomputation at init.
- Consistency with DPO: aligned method order and signatures, tokenization, _prepare_dataset, PEFT handling, ref-model preparation for distributed training, and config layout — plus a new DataCollatorForKTO and output format. Metrics moved into _compute_loss and simplified to direct averages via the shared _metrics attribute.
- Removing legacy baggage: dropped encoder-decoder support, BOS/EOS handling, null_ref_context, generate_during_eval, model_init, preprocess_logits_for_metrics, model/ref adapter names, and several dead config knobs.
- Coverage: a full test suite mirroring DPO, text collator tests, VLM tests, and slow tests.
- The promotion itself: the experimental → stable move (#6175) and shim cleanup (#6287), handled so downstream users get a clean deprecation path.

Honestly, this has been one of the more complex tasks I've taken on since joining the team, not because any single change was hard, but because it demanded sustained consistency across a ~2,000-line trainer, with every branch, comment, and edge case kept in lockstep with DPO.

Huge thanks to everyone who reviewed along the way (especially @qgallouedec ), the incremental review cadence is exactly what kept this maintainable.

KTO now sits on equal footing with our other flagship trainers. 🚀
  • 2 replies
·
sergiopaniego 
posted an update 2 months ago