Cheng Qian

Test-time harnesses can nearly double the performance of weaker models, transforming how we think about capability transfer in AI.

Co-evolution in agentic systems could unlock unprecedented levels of adaptability, allowing agents to evolve beyond human-imposed limitations.

Agents can now make more accurate decisions by effectively compressing multimodal memory, closing the gap with human performance in complex environments.

The best-informed LLM agent can absorb nearly the entire wealth pool in a coupled economy, revealing stark limitations in our understanding of agent dynamics.

MS-Resampler boosts multimodal reasoning capabilities while maintaining efficiency, outperforming traditional methods with minimal overhead.

Current visual world models show a dramatic decline in performance when faced with unconventional and impossible physical interactions, highlighting a critical gap in their generalization capabilities.

Agents struggle to maintain planning accuracy in complex tool ecosystems, with GPT-5.4's performance plummeting from 51.90% to 11.36% under severe blocking conditions.

Agents can boost their task completion rates by over 20% simply by grounding their actions in observed context rather than assumptions.

Rationale-based fine-tuning may actually undermine clinical prediction accuracy, challenging the belief that teaching models "why" can enhance their performance.

LLMs struggle with adaptive planning, achieving only 67.75% accuracy when faced with progressively revealed world and user constraints.

LMMs can't MacGyver their way out of a paper bag: they struggle to creatively repurpose objects in visually complex environments, revealing a critical gap in grounded reasoning beyond pattern recognition.