# Adaptive Reasoning

3 related articles

Subscribe to updates

NVIDIA Technical Blog Nov 11, 2026
NVIDIA squeezes a driving reasoning model onto onboard silicon: closed-loop RL lifts driving performance 48.9%, with 160–190 ms pixel-to-trajectory latency
NVIDIA released Alpamayo 2 Edge, an edge-oriented reasoning vision-language-action model that pairs a 2B-parameter Nemotron reasoning backbone with a diffusion action expert to produce Chain-of-Causation reasoning and future trajectories on DRIVE AGX Thor, generating reasoning only when needed. Closed-loop reinforcement learning post-training in AlpaGym improves closed-loop driving performance by 48.9% over the best-performing checkpoint without closed-loop RL, and the optimized model reaches 160–190 ms pixel-to-trajectory inference latency depending on the amount of reasoning.
NVIDIA released Alpamayo 2 Edge, an edge-oriented reasoning vision-language-action model that pairs a 2B-parameter Nemotron reasoning backbone with a diffusion action expert to produce Chain-of-Causation reasoning and future trajectories on DRIVE AGX Thor, generating reasoning only when needed. Closed-loop reinforcement learning post-training in AlpaGym improves closed-loop driving performance by 48.9% over the best-performing checkpoint without closed-loop RL, and the optimized model reaches 160–190 ms pixel-to-trajectory inference latency depending on the amount of reasoning.
NVIDIA
Alpamayo 2 Edge AlpaJudge Edge Deployment Adaptive Reasoning
arXiv Nov 10, 2026
Video models learn to check the evidence before reasoning: EVAThink hits 63.5% average at 32 frames with 6.7K samples and about 62% shorter output than always reasoning
EVAThink turns whether a sampled set of frames is enough to answer directly into a training signal: for the same question and answer it builds evidence-dense and evidence-sparse views of a video and keeps only pairs where direct answering degrades on the sparse view but explicit reasoning recovers accuracy, then uses two-stage reinforcement learning to teach the model to switch between direct answering and reasoning. Trained on only 6.7K samples, it reaches 63.5% and 65.6% average across six video understanding benchmarks under fixed 32- and 64-frame budgets, both above the Qwen3-VL-8B-Instruct base; at 64 frames the adaptive mode beats direct-only and think-only by 2.3 and 3.6 percentage points and cuts output length by about 62% relative to think-only.
EVAThink turns whether a sampled set of frames is enough to answer directly into a training signal: for the same question and answer it builds evidence-dense and evidence-sparse views of a video and keeps only pairs where direct answering degrades on the sparse view but explicit reasoning recovers accuracy, then uses two-stage reinforcement learning to teach the model to switch between direct answering and reasoning. Trained on only 6.7K samples, it reaches 63.5% and 65.6% average across six video understanding benchmarks under fixed 32- and 64-frame budgets, both above the Qwen3-VL-8B-Instruct base; at 64 frames the adaptive mode beats direct-only and think-only by 2.3 and 3.6 percentage points and cuts output length by about 62% relative to think-only.
EVAThink Adaptive Reasoning Video Understanding Reinforcement Learning GRPO
arXiv Oct 12, 2026
Audio reasoning stops spending equal tokens on every step: self-rated section difficulty lifts MMAR to 62.0% while cutting tokens by about 60%
Large audio-language models that reason explicitly have mostly used one overall difficulty signal to decide whether to think longer, which gives little guidance on how much effort each part of the reasoning trace deserves. SWEAR scores difficulty along five dimensions covering acoustic quality, acoustic complexity and reasoning demand, uses those offline scores to compress reference traces, teaches the model during supervised fine-tuning to predict each section's difficulty before writing it, and then lets reinforcement learning set section length from those self-ratings. On MMAU-test-mini and MMAR, accuracy rises by 2.6 and 6.3 percentage points over CoTA-SFT without difficulty supervision, while generated tokens fall by 60.1% and 62.3%.
Large audio-language models that reason explicitly have mostly used one overall difficulty signal to decide whether to think longer, which gives little guidance on how much effort each part of the reasoning trace deserves. SWEAR scores difficulty along five dimensions covering acoustic quality, acoustic complexity and reasoning demand, uses those offline scores to compress reference traces, teaches the model during supervised fine-tuning to predict each section's difficulty before writing it, and then lets reinforcement learning set section length from those self-ratings. On MMAU-test-mini and MMAR, accuracy rises by 2.6 and 6.3 percentage points over CoTA-SFT without difficulty supervision, while generated tokens fall by 60.1% and 62.3%.
Shanghai Jiao Tong University
Yixuan Li（First author）
SWEAR DiCoTA Adaptive Reasoning GRPO Large Audio Language Models

You’re all caught up

NVIDIA squeezes a driving reasoning model onto onboard silicon: closed-loop RL lifts driving performance 48.9%, with 160–190 ms pixel-to-trajectory latency

Video models learn to check the evidence before reasoning: EVAThink hits 63.5% average at 32 frames with 6.7K samples and about 62% shorter output than always reasoning

Audio reasoning stops spending equal tokens on every step: self-rated section difficulty lifts MMAR to 62.0% while cutting tokens by about 60%