Blog - paper-notes

ConceptSeg-R1: teaching a segmenter to induce the rule, not just solve the instance
Paper notes

Paper notes on ConceptSeg-R1 — RL-style reasoning over SAM 3's prompt space, using Meta-GRPO to infer transferable segmentation rules from a few visual demonstrations.

paper-notes segmentation reinforcement-learning sam

LocateAnything: decoding bounding boxes in parallel instead of one token at a time
Paper notes

Paper notes on NVIDIA's LocateAnything — a 3B vision-language grounding model that treats boxes as atomic units via Parallel Box Decoding, hitting 10× the throughput of comparable VLMs.

paper-notes vision-language grounding nvidia

QuCo-RAG: deciding when to retrieve by asking the pre-training corpus
Paper notes

Paper notes on QuCo-RAG — using entity frequency and co-occurrence in a 4T-token corpus as a retrieval trigger, instead of the model's own unreliable uncertainty.

rag paper-notes llm uncertainty

ConceptSeg-R1: teaching a segmenter to induce the rule, not just solve the instance Paper notes

LocateAnything: decoding bounding boxes in parallel instead of one token at a time Paper notes

QuCo-RAG: deciding when to retrieve by asking the pre-training corpus Paper notes

ConceptSeg-R1: teaching a segmenter to induce the rule, not just solve the instance
Paper notes

LocateAnything: decoding bounding boxes in parallel instead of one token at a time
Paper notes

QuCo-RAG: deciding when to retrieve by asking the pre-training corpus
Paper notes