The success of large-scale pretraining in NLP and computer vision has catalyzed growing efforts to develop analogous foundation models for the physical sciences. However, pretraining strategies using atomistic data remain underexplored. To date, large-scale supervised pretraining on DFT force-energy labels has provided the strongest performance gains to downstream property prediction, out-performing existing methods of self-supervised learning (SSL) which remain limited to ground-state geometries, and/or single domains of atomistic data. We address these shortcomings with Self-Conditioned Denoising (SCD), a […]

Ver mais

Humanoid robots can suffer significant performance drops under small changes in dynamics, task specifications, or environment setup. We propose HoRD, a two-stage learning framework for robust humanoid control under domain shift. First, we train a high-performance teacher policy via history-conditioned reinforcement learning, where the policy infers latent dynamics context from recent state–action trajectories to adapt online to diverse randomized dynamics. Second, we perform online distillation to transfer the teacher’s robust control capabilities into a transformer-based student policy that […]

Ver mais

Over the last couple of months I built a full LLM training pipeline from scratch in PyTorch architecture, pretraining, SFT, reward modeling, and three post-training alignment methods. No pretrained weights, no alignment libraries. I just published the final comparison study. The short version: Phase 1 results (baseline hyperparameters): PPO: +3.99 → GRPO: -0.12 → DPO: +2.40 (average reward on 16 fixed prompts) Phase 5 results (after targeted tuning): DPO: +4.15 → SFT: +4.13 → GRPO: +3.31 → PPO: […]

Ver mais

arXiv:2601.15498v1 Announce Type: new Abstract: Speculative Decoding (SD) accelerates autoregressive large language model (LLM) inference by decoupling generation and verification. While recent methods improve draft quality by tightly coupling the drafter with the target model, the verification mechanism itself remains largely unchanged, relying on strict token-level rejection sampling. In practice, modern LLMs frequently operate in low-margin regimes where the target model exhibits weak preference among top candidates. In such cases, rejecting plausible runner-up tokens yields negligible information gain […]

Ver mais

arXiv:2510.23254v2 Announce Type: replace Abstract: We study in-context learning problems where a Transformer is pretrained on tasks drawn from a mixture distribution $pi=sum_{alphainmathcal{A}} lambda_{alpha} pi_{alpha}$, called the pretraining prior, in which each mixture component $pi_{alpha}$ is a distribution on tasks of a specific difficulty level indexed by $alpha$. Our goal is to understand the performance of the pretrained Transformer when evaluated on a different test distribution $mu$, consisting of tasks of fixed difficulty $betainmathcal{A}$, and with potential distribution […]

Ver mais