Tag: post-training
Browse our exclusive articles!
Boost LLM Performance with Multilingual Post-Training
Explore how multilingual post-training improves large language models, enhancing low-resource languages and cross-lingual transfer beyond English-only trai...
Emergent Attention Heads in Post-Training Reasoning Models
Discover how post-training sparks emergent attention heads that enhance reasoning models' performance and reveal key trade-offs in AI computation.
Lightning OPD: Fast Offline Distillation for Large Models
Discover Lightning OPD, an efficient offline on-policy distillation method that boosts large reasoning model training without live servers.
SOAR: Enhancing Diffusion Models with Self-Correction
Discover SOAR, a post-training method that boosts diffusion model performance via self-correction, improving alignment and refinement without reward models...
TorchUMM: Unified Codebase for Multimodal Model Evaluation
Discover TorchUMM, the unified codebase for evaluating, analyzing, and post-training diverse multimodal AI models across visual and textual data.