…To put it directly, RL fundamentally optimizes the recall of latent knowledge. 2️⃣ The unexpected contribution of 0/128 samples: Remarkably, ~83% of the performance jump is driven by training on the…
…These developments lead to strong performance in multimodal coding, visual tool use , and framework-based agentic tasks, while preserving competitive text-only coding capability. More importantly, our development process offers practical insights…
…Enhancing Group Relative Policy Optimization with Verifiable Process Supervision for Effective Reasoning (2026) Revisiting On-Policy Distillation: Empirical Failure Modes and Simple Fixes (2026) Self-Induced Outcome Potential: Turn-Level Credit Assignment…
…The queries are optimized end-to-end with the video diffusion transformers ( DiTs ), forming an evolving memory that supports arbitrary compression ratios with constant computation independent of video length. They also act…
…In this work, we empirically find that mandating explicit object boxes in visually grounded CoT during inference often degrades performance compared to standard textual CoT, which reasons without explicit visual grounding. We…
…Comparable Performance on End-Side Devices (2026) $\phi$-Balancing for Mixture-of-Experts Training (2026) Post-Trained MoE Can Skip Half Experts via Self-Distillation (2026) DOT-MoE: Differentiable Optimal Transport for…
…We present RewardHarness, a self-evolving agentic reward framework that reframes reward modeling as context evolution rather than weight optimization. Instead of learning from large-scale annotations, RewardHarness aligns with human preferences…
…Yu-Hsi Chen , Abstract CPCANet is a domain generalization framework that uses Common Principal Component Analysis to discover structured domain-invariant subspaces through differentiable neural layers, achieving state-of-the-art performance…
…To prevent cheating, we propose CapReward, a reward design based on the CapCode principle to discourage optimization beyond the cap. Experiments across multiple datasets show that CapCode detects cheating while preserving performance…
…Zhuoming Liu , , , , , Abstract DRIFT is a framework that adapts pretrained vision-language models for continuous decoding tasks by combining coarse prediction with iterative refinement through flow matching, improving performance across perception and…
