Yifan Yang

Standard token-level SAEs squander their sparsity budgets memorizing punctuation and syntax; training on pooled multi-token chunks forces dictionaries to capture true, steerable semantic concepts and latent reasoning.

Frontier LLMs can solve 97% of math problems yet struggle to distinguish correct answers from errors with confidence better than chance.

This work introduces CoRA-NAS (COarse Ranking + Anchor-residual), a two-stage framework combining a static ranking prior with low-cost learning-curve refinement that combines cross-space ranking robustness with low-cost architecture selection.

Urban spatial optimization can be solved far more effectively by framing it as an interactive code-and-eval sandbox with decomposed build-and-refine RL than by relying on rigid, task-specific RL solvers.

XR-2 shows that scaling demonstration data and incorporating real-time corrections can dramatically enhance bimanual manipulation success rates in household tasks.

RACE reveals that a scalable statistical approach can vastly improve our understanding of neuron behavior in LLMs while slashing computational costs.

Current vision-language models falter in 3D assembly reasoning, struggling with accuracy and complexity in real-world industrial scenarios.

Verifiable temporal grounding in video forensics can drastically improve the detection of AI-generated content, outperforming traditional methods reliant on coarse supervision.

Current LLMs fail to meet the rigorous demands of PCB routing, showing major weaknesses in path planning and constraint adherence.

Illumination variations can destabilize spacecraft pose estimation, but PAID-ViT achieves remarkable robustness by disentangling illumination effects from structural cues.

Current models struggle with hybrid interface tasks, achieving only a 41.2% success rate, underscoring a critical gap in CUA evaluation.