P. Torr

LLMs exhibit "Thinking Inertia," stubbornly continuing to show explicit reasoning even when explicitly instructed not to, particularly in open-ended tasks.

Visual representations of graphs can transform how models reason and learn, bridging a critical gap between symbolic and visual understanding in AI.

Models trained with VBVR-Pro not only excel in a controlled task space but also show significant transferability to external benchmarks, revealing critical insights into visual reasoning mechanisms.

MMDiff reveals that multimodal SAEs can be powerful tools for both auditing and steering MLLM behavior, achieving up to 24% reduction in safety attack success rates without compromising performance on visual question answering.

Early unification in multimodal training can prevent models from becoming overly reliant on language, unlocking new efficiencies in generative performance.

AI scientists excel at idea generation but falter in filtering and prioritizing innovations, revealing a critical gap in their capabilities.

Adversarial attacks on vision-language agents reveal critical vulnerabilities, with multi-view optimization strategies proving significantly more effective than isolated approaches.

SciReasoner achieves a remarkable 31% improvement in Cellular Component annotation for low-homology proteins, showcasing the power of structural reasoning in AI.