Tao Yang

This paper introduces PASS (Process Advantage Signal Shaping), a method to address three pathologies in GRPO (Group Relative Policy Optimization) for process-supervised reinforcement learning of LLM reasoners.

This paper proposes Trellis, a data foundation that treats experience graphs from long-horizon agentic tasks as first-class, governed, queryable database state.

This paper proposes a scheme to optimize model sparsity in neural sparse retrieval using adaptive strategies.

This paper proposes GUMPIM, a PIM system that enables interleaved CPU and PIM pages to coexist in a single memory space.