paper
AIR: Adaptive Interleaved Reasoning with Code in MLLMs
paperactiveprovisional
air-adaptive-interleaved-reasoning-with-code-in-mllms-d55836ea·1 events·first seen 40h agoAliases: AIR: Adaptive Interleaved Reasoning with Code in MLLMs
Co-occurring entities
More like this (12)
Watch, Remember, Reason: Human-View Video Understanding with MLLMsAgentic Chain-of-Thought Steering for Efficient and Controllable LLM ReasoningAdaptive Parallel ReasoningContinual LLM Upcycling: A Predictor-Gated Bank-Wise Sparsity Training Recipe for Dense-to-Sparse LLMsReasoning in Memory (RiM)Towards Root Memories: Benchmarking and Enhancing Implicit Logical Memory Retrieval for Personalized LLMsCLP: Collocation-Length Prediction for Zero-Loss Adaptive Multi-Token Inferencecode synthesis LLMsWhere Does the Answer Come From? Benchmarking View-Level Visual Evidence Identification in Multi-View MLLMs for Autonomous DrivingExpRL: Exploratory RL for LLM Mid-TrainingMLSkip: Data Skipping for ML Filters via Lightweight MetadataTensorRT-LLM
Recent events (1)
AIR: Adaptive Interleaved Reasoning with Code in Multimodal LLMs via Reinforcement Learning
Researchers propose AIR, a system that trains multimodal large language models to adaptively interleave reasoning with code execution for numerical computation tasks, going beyond prior work that focused only on visual operations. The approach combines a two-stage cold-start data pipeline, RL dataset filtering, and a group-constrained reward function for tool-invocation decisions. Experiments show a 6.1 percentage point average improvement on evaluation benchmarks, with interleaved reasoning samples gaining 9.9 pp and tool-use success exceeding 95%.