Official Implementation of Gamer: A Generalist Vision-Language-Action Foundation Model for Long-Horizon Real-Time Game Agents
Adapting Qwen2.5-14B to Brazilian SUS clinical guidelines. Includes 2 open benchmarks (HealthBench-BR, PCDT-QA) and 8 model checkpoints from the paper's ablations.
Data Preparation for Large Language Models — a curated companion to our JCST 2026 survey. Covers Pre-training, Continual Pre-training, and Post-training (SFT/RLHF/RLAIF) across collection, filtering, dedup, generation, evaluation.
Open research in embodied AI and cognitive architectures, investigating autonomous learning from first-person experience through reproducible experiments.
End-to-end 35.5M-parameter small language model in medical domain built from scratch in PyTorch with retention-aware continual pretraining, masked SFT, exact resume, and sealed evaluation.
An industry-grade visual studio for training custom subword tokenizers (BBPE, Unigram, WordPiece), safely merging them into frontier LLMs (Qwen 2.5/3, LLaMA 3, DeepSeek) with zero base regression, and initializing Tensor Core-aligned embeddings for Continued Pre-Training (CPT).
Continual Learning done simplistic
From-scratch JAX/Flax language-model training lab for controlled continual pretraining, replay, and reproducible evaluation experiments.
To associate your repository with the continual-pretraining topic, visit your repo's landing page and select "manage topics."