Search: “safety”

Response Time Enhances Alignment with Heterogeneous Preferences

arXiv:2605.06987v1 Announce Type: new Abstract: Aligning large language models (LLMs) to human preferences typically relies on aggregating pooled feedback into a single reward model. However, this sta

Rethinking Importance Sampling in LLM Policy Optimization: A Cumulative Token Perspective

DGX agent

arXiv:2605.07331v1 Announce Type: cross Abstract: Reinforcement learning, including reinforcement learning with verifiable rewards (RLVR), has emerged as a powerful approach for LLM post-training. Cen

safetyarxiv-cs-ai

11 May 2026

RIDER: 3D RNA Inverse Design with Reinforcement Learning-Guided Diffusion

arXiv:2602.16548v2 Announce Type: replace Abstract: The inverse design of RNA three-dimensional (3D) structures is crucial for engineering functional RNAs in synthetic biology and therapeutics. While

Risk-Consistent Multiclass Learning from Random Label-Subset Membership Queries

arXiv:2605.07413v1 Announce Type: new Abstract: Obtaining accurate class labels is often costly or unreliable, and may also be limited by privacy or other practical conditions. Compared with asking an

Robustness of Refugee-Matching Gains to Off-Policy Evaluation Choices

arXiv:2605.06686v1 Announce Type: new Abstract: Previous research has investigated the potential of refugee matching for boosting refugee outcomes, first considered by Bansak et al. (2018). This paper

Rollback-Free Stable Brick Structures Generation

arXiv:2605.06947v1 Announce Type: new Abstract: While autoregressive models have advanced 3D generation, creating physically stable brick structures remains a challenge due to the strict requirements

Rubric-based On-policy Distillation

arXiv:2605.07396v1 Announce Type: cross Abstract: On-policy distillation (OPD) is a powerful paradigm for model alignment, yet its reliance on teacher logits restricts its application to white-box sce

Safactory: A Scalable Agentic Infrastructure for Training Trustworthy Autonomous Intelligence

arXiv:2605.06230v2 Announce Type: replace Abstract: As large models evolve from conversational assistants into autonomous agents, challenges increasingly arise from long-horizon decision making, tool

SAGE: Hierarchical LLM-Based Literary Evaluation through Ontology-Grounded Interpretive Dimensions

arXiv:2605.07102v1 Announce Type: new Abstract: Evaluating literary quality requires assessing interpretive dimensions such as cultural representation, emotional depth, and philosophical sophisticatio

Same Signal, Opposite Meaning: Direction-Informed Adaptive Learning for LLM Agents

arXiv:2605.06908v1 Announce Type: cross Abstract: Adaptive test-time compute for LLM agents aims to invoke extra computation only when it improves performance. Existing methods typically use confidenc

SARA: Semantically Adaptive Relational Alignment for Video Diffusion Models

arXiv:2605.07800v1 Announce Type: new Abstract: Recent video diffusion models (VDMs) synthesize visually convincing clips, yet still drop entities, mis-bind attributes, and weaken the interactions spe

Science publishing giant Elsevier has joined the dozens of firms and individuals suing artificial intelligence companies over their alleged …

Science publishing giant Elsevier has joined the dozens of firms and individuals suing artificial intelligence companies over their alleged use of copyrighted works in training AI models https://go.na

Self-Programmed Execution for Language-Model Agents

arXiv:2605.06898v1 Announce Type: new Abstract: At the heart of existing language model agents is a fixed orchestrator program responsible for the state transition between consecutive turns. This pape

Serious question: Should I write a short book called 7 lies about AI that never die?

Serious question: Should I write a short book called 7 lies about AI that never die? AI hype has become a giant game of bait and switch. the bait: we are going to make an AI that can solve any problem

SHARP: A Self-Evolving Human-Auditable Rubric Policy for Financial Trading Agents

arXiv:2605.06822v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly deployed for autonomous financial trading, a domain requiring continuous adaptation to noisy, non-stationa

SimCT: Recovering Lost Supervision for Cross-Tokenizer On-Policy Distillation

arXiv:2605.07711v1 Announce Type: new Abstract: On-policy distillation (OPD) is a standard tool for transferring teacher behavior to a smaller student, but it implicitly assumes that teacher and stude

Since Hinton has actually replied let me clarify some things - LLMS don’t always regurgitate - LLMs don’t literally store full texts - but…

Since Hinton has actually replied let me clarify some things - LLMS don’t *always* regurgitate - LLMs don’t literally store full texts - but given the mechanisms that they use they do sometimes regurg

Skill1: Unified Evolution of Skill-Augmented Agents via Reinforcement Learning

arXiv:2605.06130v2 Announce Type: replace Abstract: A persistent skill library allows language model agents to reuse successful strategies across tasks. Maintaining such a library requires three coupl

Slowly Annealed Langevin Dynamics: Theory and Applications to Training-Free Guided Generation

arXiv:2605.07950v1 Announce Type: new Abstract: We study Slowly Annealed Langevin Dynamics (SALD), a sampler for tracking a path of moving target distributions and approximating the terminal target th

SOD: Step-wise On-policy Distillation for Small Language Model Agents

arXiv:2605.07725v1 Announce Type: cross Abstract: Tool-integrated reasoning (TIR) is difficult to scale to small language models due to instability in long-horizon tool interactions and limited model

Sources: the White House's Office of the National Cyber Director and Commerce Department's CAISI are fighting over which agency should lead AI model evaluations (Washington Post)

Washington Post: Sources: the White House's Office of the National Cyber Director and Commerce Department's CAISI are fighting over which agency should lead AI model evaluations — As the White House g

safetytechmeme

SparseRL-Sync: Lossless Weight Synchronization with ~100x Less Communication

arXiv:2605.07330v1 Announce Type: cross Abstract: In large-scale reinforcement learning (RL) systems with decoupled Trainer-Rollout execution, the Trainer must regularly synchronize policy weights to

Stabilized neural Hamilton--Jacobi--Bellman solvers: Error analysis and applications in model-based reinforcement learning

arXiv:2605.07116v1 Announce Type: cross Abstract: Physics-informed neural solvers offer a promising route to model-based reinforcement learning in continuous time, where optimal feedback synthesis is

Stationary Reweighting Yields Local Convergence of Soft Fitted Q-Iteration

arXiv:2512.23927v2 Announce Type: replace-cross Abstract: Fitted Q-iteration (FQI) and soft FQI are widely used value-based methods for offline reinforcement learning, but their standard stability gua

STDA-Net: Spectrogram-Based Domain Adaptation for cross-dataset Sleep Stage Classification

arXiv:2605.06736v1 Announce Type: cross Abstract: Accurate sleep stage classification across datasets remains challenging due to variability in EEG channel montages, sampling rates, recording environm

Structured Role-Aware Policy Optimization for Multimodal Reasoning

arXiv:2605.07274v1 Announce Type: new Abstract: Reinforcement learning from verifiable rewards (RLVR), especially with Group Relative Policy Optimization (GRPO), has shown strong potential for improvi

Supervised sparse auto-encoders for interpretable and compositional representations

arXiv:2602.00924v2 Announce Type: replace Abstract: Sparse auto-encoders (SAEs) have re-emerged as a prominent method for mechanistic interpretability, yet they face two significant challenges: the no

Temporal Attention for Adaptive Control of Euler-Lagrange Systems with Unobservable Memory

arXiv:2605.06877v1 Announce Type: new Abstract: Adaptive control of Euler-Lagrange systems is challenging when friction is governed by a finite-horizon internal state that is not directly observable f

Temporal Smoothness Doubly Robust Learning for Debiased Knowledge Tracing

arXiv:2605.05958v2 Announce Type: replace Abstract: Knowledge Tracing (KT) is fundamental to intelligent education systems, yet relies on educational logs that are selectively observed. The non-random

TextLDM: Language Modeling with Continuous Latent Diffusion

arXiv:2605.07748v1 Announce Type: new Abstract: Diffusion Transformers (DiT) trained with flow matching in a VAE latent space have unified visual generation across images and videos. A natural next st

That @NIST CAISI announcement from last week about Google DeepMind, xAI and Microsoft signing new deals for pre-deployment testing is gone f…

That @NIST CAISI announcement from last week about Google DeepMind, xAI and Microsoft signing new deals for pre-deployment testing is gone from their website, link no longer found, and I can't get any

The Cost of Consensus: Malignant Epistemic Herding and Adaptive Gating in Distributed Multi-Agent Search

arXiv:2605.06988v1 Announce Type: cross Abstract: Distributed agents in real-world settings frequently must coordinate under uncertainty with only partial observations. Coordination is necessary to sh

safetyarxiv-cs-ro

The Effect of Mini-Batch Noise on the Implicit Bias of Adam

arXiv:2602.01642v2 Announce Type: replace-cross Abstract: With limited high-quality data and growing compute, multi-epoch training is gaining back its importance across sub-areas of deep learning. Ada

The Endogeneity of Miscalibration: Impossibility and Escape in Scored Reporting

arXiv:2605.07671v1 Announce Type: cross Abstract: Eliciting truthful reports from autonomous agents is a core problem in scalable AI oversight: a principal scores the agent's report using a strictly p

The Limits of AI-Driven Allocation: Optimal Screening under Aleatoric Uncertainty

arXiv:2605.07979v1 Announce Type: new Abstract: The rise of machine learning has shifted targeted resource allocation in policy and humanitarian settings toward algorithmic targeting based on predicte

The Proxy Presumption: From Semantic Embeddings to Valid Social Measures

arXiv:2605.07409v1 Announce Type: new Abstract: Natural Language Processing is rapidly evolving into a primary instrument for Computational Social Science, with researchers increasingly using embeddin

Think-with-Rubrics: From External Evaluator to Internal Reasoning Guidance

arXiv:2605.07461v1 Announce Type: new Abstract: Rubrics have been extensively utilized for evaluating unverifiable, open-ended tasks, with recent research incorporating them into reward systems for re

This is what a useless hype lifecycle looks like.

Gary Marcus critiques the typical hype cycle pattern where emerging technologies experience inflated expectations followed by inevitable disappointment. The post likely illustrates this cycle using a

totally worth $10 trillion a year

This post from AI researcher Gary Marcus likely discusses the enormous economic value or potential return on investment related to artificial intelligence developments, suggesting AI's worth or impact

Toward Better Geometric Representations for Molecule Generative Models

arXiv:2605.07693v1 Announce Type: new Abstract: Geometric representation-conditioned molecule generation provides an effective paradigm that decouples molecule representation modeling from structure g

Towards Differentially Private Reinforcement Learning with General Function Approximation

arXiv:2605.07049v1 Announce Type: cross Abstract: We present the first theoretical guarantees for differentially private online reinforcement learning (RL) with general function approximation, extendi

Towards Fairness under Label Bias in Image Segmentation: Impact, Measurement and Mitigation

arXiv:2605.06891v1 Announce Type: new Abstract: Labeled datasets reflect the biases of their annotation pipelines, which sometimes introduce label bias: group-conditional label errors that cause syste

TRACE: Transport Alignment Conformal Prediction via Diffusion and Flow Matching Models

arXiv:2605.07100v1 Announce Type: cross Abstract: Constructing valid and informative conformal prediction regions for multi-dimensional outputs remains a fundamental challenge. While conformal predict

Training-Free Multimodal Large Language Model Orchestration

arXiv:2508.10016v3 Announce Type: replace Abstract: Building interactive omni-modal assistants often relies on end-to-end multimodal alignment to fuse heterogeneous modalities, which incurs substantia

TRAJGANR: Trajectory-Centric Urban Multimodal Learning via Geospatially Aligned Neural Representations

arXiv:2605.06990v1 Announce Type: new Abstract: Multimodal self-supervised learning (MSSL) has emerged as a key paradigm for pretraining geospatial foundation models. However, existing geospatial MSSL

UFT: Unifying Fine-Tuning of SFT and RLHF/DPO/UNA through a Generalized Implicit Reward Function

arXiv:2410.21438v3 Announce Type: replace Abstract: By pretraining on trillions of tokens, an LLM gains the capability of text generation. However, to enhance its utility and reduce potential harm, SF

UNA: A Unified Supervised Framework for Efficient LLM Alignment Across Feedback Types

arXiv:2408.15339v4 Announce Type: replace-cross Abstract: RL alignment methods, including RLHF and DPO, are primarily based on pairwise preference data. Although scalar or score-based feedback has bee

UniD-Shift: Towards Unified Semantic Segmentation via Interpretable Share-Private Multimodal Decomposition

arXiv:2605.07356v1 Announce Type: new Abstract: Semantic segmentation of large-scale 3D point clouds is crucial for applications such as autonomous driving and urban digital twins. However, the sparse