2 papers
cs.LG2026
Divide and Cooperate: Role-Decomposed Multi-Agent LLM Training with Cross-Agent Learning Signals
Jaewan Park, Solbee Cho, Jay-Yoon Lee
Modern language agents which perform multi-step reasoning have shown strong performance in knowledge-intensive question answering. However, existing approaches typically couple evi…
cs.AI2026
Shallow Prefill, Deep Decoding: Efficient Long-Context Inference via Layer-Asymmetric KV Visibility
Jungsuk Oh, Hyeseo Jeon, Hyunjune Ji +2
Long-context inference in decoder-only language models is costly because long prompts are processed during Prefill, cached at every layer, and repeatedly attended to during autoreg…