2 papers
cs.LG2026
SPARK: Learning from Verification in Speculative GRPO Rollouts
Nhat Minh Pham, Duy Tung Doan, Thi Duyen Ngo +3
Group Relative Policy Optimization (GRPO) has become a standard approach for improving the reasoning abilities of large language models, but its training is costly because each ste…
cs.SE2026
CollabCoder: Plan-Code Co-Evolution via Collaborative Decision-Making for Efficient Code Generation
Duy Tung Doan, Quang Huy Phung, Dzung Nguyen +1
Automated code generation remains a persistent challenge in software engineering, as conventional multi-agent frameworks are often constrained by static planning, isolated executio…