Topic
theory
Thermodynamic Measure of Intelligence Proposed: Rare-Valid Futures Amplification as Universal Scale
A new theoretical framework proposes intelligence as the lawful amplification of rare but valid futures. The research presents a thermodynamic measure and shows that recursive self-simulation is necessary and nearly sufficient for high thermodynamic intelligence. This makes intelligence measurable on a universal scale.
New Research Shows Chain-of-Thought Reasoning Should Be Selective, Not Default, for LLMs
A research paper on arXiv argues that chain-of-thought (CoT) reasoning should not be the default for large language models. The authors propose EDRM, a training-free routing framework that uses early decoding entropy to decide when to use CoT, achieving up to 55% token reduction and accuracy improvements across 15 benchmarks.
New Generalization Bounds for Deep Learning Models via Local Robustness and Stability
Researchers propose a new generalization bound for deep learning models that accounts for local variation in robustness across input sub-regions. Experiments on ImageNet show the bounds are non-vacuous and tighter than existing methods, aligning closely with empirical performance.