#grpo
Grpo: 7 AI articles covering grpo news, analysis, and research
Articles
Can an Open Model Do Security Research? Cantina's apex-flash-1⭐7
Cantina's open-weights apex-flash-1, a fine-tuned GLM-5.3-Flash model, solves 40 of 60 held-out bug tasks for vulnerability research.
Improving OCR Faithfulness via Gated and Attenuated On-Policy⭐8
GAD-RL adaptively gates and attenuates teacher supervision during on-policy distillation, improving OCR faithfulness and achieving strong CHAOS-Bench and OmniDo...
ChestPheNoT: Deployable, Auditable Label-Status-Evidence⭐9
ChestPheNoT extracts finding labels, three-class status, and verbatim evidence spans from radiology reports using a locally deployable, auditable 0.5–3B model.
Fine-tuning a 350M Model for Better Structured Outputs in 100⭐9
Fine-tuning a 350M model with 100 GRPO steps boosts structured output accuracy, cutting syntax errors by 35% for efficient AI deployment.
AllenAI Open Instruct Tulu 3: Building a Compact Post-Training⭐9
Learn to build a compact post-training pipeline with AllenAI's Tulu 3: SFT, DPO, and RLVR using GRPO in a single-GPU environment.
Be Consistent! Enhancing Robust Visual Reasoning in LVLMs with⭐7
New benchmark ConVBench and training method ConVLM enhance logical consistency in vision-language AI, evaluating and improving robust visual reasoning.
Adversarial Style Optimization: Enhancing VLM Jailbreaks by⭐7
ASO enhances visual jailbreak attacks on multimodal LLMs by optimizing stylistic triggers via GRPO, achieving higher Attack Success Rates and exposing stylistic...