ML algorithms, training, and inference
One production problem that feels bigger than people admit: a model looks fine, sounds safe, and then gives away too much the moment some...
One production problem that feels bigger than people admit: a model looks fine, sounds safe, and then gives away too much the moment some...
So, yesterday run was a success and I did get an avg rollout length of about 64 tokens as attached in the image! This was with quality_re...
![[2603.28248] Reasoning as Energy Minimization over Structured Latent Trajectories](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
Abstract page for arXiv paper 2603.28248: Reasoning as Energy Minimization over Structured Latent Trajectories
Abstract page for arXiv paper 2603.28197: EpiPersona: Persona Projection and Episode Coupling for Pluralistic Preference Modeling
Abstract page for arXiv paper 2603.28183: PReD: An LLM-based Foundation Multimodal Model for Electromagnetic Perception, Recognition, and...
Abstract page for arXiv paper 2603.28135: CoT2-Meta: Budgeted Metacognitive Control for Test-Time Reasoning
Abstract page for arXiv paper 2603.28062: SLOW: Strategic Logical-inference Open Workspace for Cognitive Adaptation in AI Tutoring
Abstract page for arXiv paper 2603.28052: Meta-Harness: End-to-End Optimization of Model Harnesses
Abstract page for arXiv paper 2603.28026: When Choices Become Priors: Contrastive Decoding for Scientific Figure Multiple-Choice QA
Abstract page for arXiv paper 2603.28015: What an Autonomous Agent Discovers About Molecular Transformer Design: Does It Transfer?
Abstract page for arXiv paper 2603.28010: HeteroHub: An Applicable Data Management Framework for Heterogeneous Multi-Embodied Agent System
Abstract page for arXiv paper 2603.27977: SARL: Label-Free Reinforcement Learning by Rewarding Reasoning Topology
Abstract page for arXiv paper 2603.27958: CARV: A Diagnostic Benchmark for Compositional Analogical Reasoning in Multimodal LLMs
Abstract page for arXiv paper 2603.27751: SkyNet: Belief-Aware Planning for Partially-Observable Stochastic Games
Abstract page for arXiv paper 2603.27738: TianJi:An autonomous AI meteorologist for discovering physical mechanisms in atmospheric science
Abstract page for arXiv paper 2603.27438: The Novelty Bottleneck: A Framework for Understanding Human Effort Scaling in AI-Assisted Work
Abstract page for arXiv paper 2603.27423: AstraAI: LLMs, Retrieval, and AST-Guided Assistance for HPC Codebases
Abstract page for arXiv paper 2603.27404: Heterogeneous Debate Engine: Identity-Grounded Cognitive Architecture for Resilient LLM-Based E...
Abstract page for arXiv paper 2603.27360: Defend: Automated Rebuttals for Peer Review with Minimal Author Guidance
Abstract page for arXiv paper 2603.27343: Beyond Completion: Probing Cumulative State Tracking to Predict LLM Agent Performance
Abstract page for arXiv paper 2603.27338: CounterMoral: Editing Morals in Language Models
Abstract page for arXiv paper 2603.27314: TokenDance: Token-to-Token Music-to-Dance Generation with Bidirectional Mamba
Get the latest news, tools, and insights delivered to your inbox.
Daily or weekly digest • Unsubscribe anytime