arXiv:2602.06107v1 Announce Type: new
Abstract: Reinforcement learning (RL) for large language models (LLMs) remains expensive, particularly because the rollout is expensive. Decoupling rollout generation from policy optimization (e.g., leveraging a more efficient model to rollout) could enable sub...
arXiv:2602.06176v1 Announce Type: new
Abstract: Large Language Models (LLMs) have exhibited remarkable reasoning capabilities, achieving impressive results across a wide range of tasks. Despite these advances, significant reasoning failures persist, occurring even in seemingly simple scenarios. To ...
Do It for HER: First-Order Temporal Logic Reward Specification in Reinforcement Learning (Extended Version)
arXiv:2602.06227v1 Announce Type: new
Abstract: In this work, we propose a novel framework for the logical specification of non-Markovian rewards in Markov Decision Processes (MDPs) with large state spaces. Our approach leverages Linear Temporal Logic Modulo Theories over finite traces (LTLfMT), a ...
Do LLMs Act Like Rational Agents? Measuring Belief Coherence in Probabilistic Decision Making
arXiv:2602.06286v1 Announce Type: new
Abstract: Large language models (LLMs) are increasingly deployed as agents in high-stakes domains where optimal actions depend on both uncertainty about the world and consideration of utilities of different outcomes, yet their decision logic remains difficult t...
Exposing Weaknesses of Large Reasoning Models through Graph Algorithm Problems
arXiv:2602.06319v1 Announce Type: new
Abstract: Large Reasoning Models (LRMs) have advanced rapidly; however, existing benchmarks in mathematics, code, and common-sense reasoning remain limited. They lack long-context evaluation, offer insufficient challenge, and provide answers that are difficult ...
Denoising diffusion networks for normative modeling in neuroimaging
arXiv:2602.04886v1 Announce Type: new
Abstract: Normative modeling estimates reference distributions of biological measures conditional on covariates, enabling centiles and clinically interpretable deviation scores to be derived. Most neuroimaging pipelines fit one model per imaging-derived phenoty...
A Causal Perspective for Enhancing Jailbreak Attack and Defense
arXiv:2602.04893v1 Announce Type: new
Abstract: Uncovering the mechanisms behind "jailbreaks" in large language models (LLMs) is crucial for enhancing their safety and reliability, yet these mechanisms remain poorly understood. Existing studies predominantly analyze jailbreak prompts by probing lat...
Momentum Attention: The Physics of In-Context Learning and Spectral Forensics for Mechanistic Interpretability
arXiv:2602.04902v1 Announce Type: new
Abstract: The Mechanistic Interpretability (MI) program has mapped the Transformer as a precise computational graph. We extend this graph with a conservation law and time-varying AC dynamics, viewing it as a physical circuit. We introduce Momentum Attention, a ...
Mind the Performance Gap: Capability-Behavior Trade-offs in Feature Steering
arXiv:2602.04903v1 Announce Type: new
Abstract: Feature steering has emerged as a promising approach for controlling LLM behavior through direct manipulation of internal representations, offering advantages over prompt engineering. However, its practical effectiveness in real-world applications rem...
DCER: Dual-Stage Compression and Energy-Based Reconstruction
arXiv:2602.04904v1 Announce Type: new
Abstract: Multimodal fusion faces two robustness challenges: noisy inputs degrade representation quality, and missing modalities cause prediction failures. We propose DCER, a
unified framework addressing both challenges through dual-stage compression and ener...
Artificial Intelligence as Strange Intelligence: Against Linear Models of Intelligence
arXiv:2602.04986v1 Announce Type: new
Abstract: We endorse and expand upon Susan Schneider's critique of the linear model of AI progress and introduce two novel concepts: "familiar intelligence" and "strange intelligence". AI intelligence is likely to be strange intelligence, defying familiar patte...
DeepRead: Document Structure-Aware Reasoning to Enhance Agentic Search
arXiv:2602.05014v1 Announce Type: new
Abstract: With the rapid progress of tool-using and agentic large language models (LLMs), Retrieval-Augmented Generation (RAG) is evolving from one-shot, passive retrieval into multi-turn, decision-driven evidence acquisition. Despite strong results in open-dom...
Evaluating Large Language Models on Solved and Unsolved Problems in Graph Theory: Implications for Computing Education
arXiv:2602.05059v1 Announce Type: new
Abstract: Large Language Models are increasingly used by students to explore advanced material in computer science, including graph theory. As these tools become integrated into undergraduate and graduate coursework, it is important to understand how reliably t...
Towards Reducible Uncertainty Modeling for Reliable Large Language Model Agents
arXiv:2602.05073v1 Announce Type: new
Abstract: Uncertainty quantification (UQ) for large language models (LLMs) is a key building block for safety guardrails of daily LLM applications. Yet, even as LLM agents are increasingly deployed in highly complex tasks, most UQ research still centers on sing...
Rethinking imitation learning with Predictive Inverse Dynamics Models
This research looks at why Predictive Inverse Dynamics Models often outperform standard Behavior Cloning in imitation learning. By using simple predictions of what happens next, PIDMs reduce ambiguity and learn from far fewer demonstrations.
The post Rethinking imitation learning with Predictive Inv...
Paza: Introducing automatic speech recognition benchmarks and models for low resource languages
Microsoft Research unveils Paza, a human-centered speech pipeline, and PazaBench, the first leaderboard for low-resource languages. It covers 39 African languages and 52 models and is tested with communities in real settings.
The post Paza: Introducing automatic speech recognition benchmarks and mo...
Understanding the Impact of Differentially Private Training on Memorization of Long-Tailed Data
arXiv:2602.03872v1 Announce Type: new
Abstract: Recent research shows that modern deep learning models achieve high predictive accuracy partly by memorizing individual training samples. Such memorization raises serious privacy concerns, motivating the widespread adoption of differentially private t...
Reversible Deep Learning for 13C NMR in Chemoinformatics: On Structures and Spectra
arXiv:2602.03875v1 Announce Type: new
Abstract: We introduce a reversible deep learning model for 13C NMR that uses a single conditional invertible neural network for both directions between molecular structures and spectra. The network is built from i-RevNet style bijective blocks, so the forward ...
arXiv:2602.03876v1 Announce Type: new
Abstract: Standard reinforcement learning from human feedback (RLHF) trains a reward model on pairwise preference data and then uses it for policy optimization. However, while reward models are optimized to capture relative preferences, existing policy optimiza...
NeuroPareto: Calibrated Acquisition for Costly Many-Goal Search in Vast Parameter Spaces
arXiv:2602.03901v1 Announce Type: new
Abstract: The pursuit of optimal trade-offs in high-dimensional search spaces under stringent computational constraints poses a fundamental challenge for contemporary multi-objective optimization. We develop NeuroPareto, a cohesive architecture that integrates ...
GeoIB: Geometry-Aware Information Bottleneck via Statistical-Manifold Compression
arXiv:2602.03906v1 Announce Type: new
Abstract: Information Bottleneck (IB) is widely used, but in deep learning, it is usually implemented through tractable surrogates, such as variational bounds or neural mutual information (MI) estimators, rather than directly controlling the MI I(X;Z) itself. T...
Knowledge Model Prompting Increases LLM Performance on Planning Tasks
arXiv:2602.03900v1 Announce Type: new
Abstract: Large Language Models (LLM) can struggle with reasoning ability and planning tasks. Many prompting techniques have been developed to assist with LLM reasoning, notably Chain-of-Thought (CoT); however, these techniques, too, have come under scrutiny as...
Enhancing Mathematical Problem Solving in LLMs through Execution-Driven Reasoning Augmentation
arXiv:2602.03950v1 Announce Type: new
Abstract: Mathematical problem solving is a fundamental benchmark for assessing the reasoning capabilities of artificial intelligence and a gateway to applications in education, science, and engineering where reliable symbolic reasoning is essential. Although r...