safety

arXiv:2604.27495v1 Announce Type: cross Abstract: Reward models (RMs) play a central role in aligning large language models (LLMs) with human preferences. However, RMs are often sensitive to spurious

arXiv:2603.09117v2 Announce Type: replace-cross Abstract: Reinforcement Learning from Verifiable Rewards (RLVR) significantly enhances large language models (LLMs) reasoning but severely suffers from

arXiv:2604.28018v1 Announce Type: cross Abstract: Design Structure Matrix (DSM) modularization, the task of partitioning system elements into cohesive modules, is a fundamental combinatorial challenge

arXiv:2604.26958v1 Announce Type: cross Abstract: Agentic AI systems capable of autonomous goal setting and proactive intervention introduce new challenges for regulating moral-emotional processes in

arXiv:2604.27045v1 Announce Type: cross Abstract: As Large Language Model (LLM) agents transition from single-session tools to persistent systems managing longitudinal healthcare journeys, their memor

arXiv:2604.27166v1 Announce Type: new Abstract: We focus on the problem of Answer-Level Fine-Tuning (ALFT), where the goal is to optimize a language model based on the correctness or properties of its

arXiv:2604.27367v1 Announce Type: cross Abstract: Simulating optical tactile sensors presents significant challenges due to their high deformability and intricate optical properties. To address these

arXiv:2604.27994v1 Announce Type: new Abstract: Learned driving agents often degrade when deployed in unseen environments. This paper studies a deliberately bounded instance of that problem in the CAR

arXiv:2604.27019v1 Announce Type: cross Abstract: Safety-aligned language models must refuse harmful requests without collapsing into broad over-refusal, but the training-time mechanisms behind this t

arXiv:2505.22798v3 Announce Type: replace-cross Abstract: The growing reliance on artificial intelligence in safety- and security-critical applications is raising concerns about the robustness of neur

This post references an interview featuring Elon Musk and Sam Altman discussing AI safety, existential risks, and the future of humanity, with an unconventional or humorous element involving goblins.

arXiv:2604.28182v1 Announce Type: cross Abstract: Reinforcement learning (RL) has become essential to the post-training of large language models (LLMs) for reasoning, agentic capabilities and alignmen

arXiv:2604.27454v1 Announce Type: new Abstract: Large language models (LLMs) sometimes exhibit qualitative shifts in response style under sustained self-referential dialogue conditions (Berg et al., 2

arXiv:2507.07986v3 Announce Type: replace-cross Abstract: We study the problem of training and fine-tuning expressive policies with online reinforcement learning (RL) given an offline dataset. Trainin

arXiv:2604.27669v1 Announce Type: new Abstract: The incorporation of fairness into the distribution network (DN) planning and operation has become a key goal of recent studies. The cost of implementin

arXiv:2604.27807v1 Announce Type: new Abstract: The design of embedded safety-critical systems such as those used in next-generation automotive and autonomous platforms, is increasingly challenged by

arXiv:2306.10407v3 Announce Type: replace-cross Abstract: Inverse reinforcement learning (IRL) is a powerful paradigm for uncovering the incentive structure that drives agent behavior, by inferring an

arXiv:2604.27875v1 Announce Type: new Abstract: AI-generated images are becoming increasingly realistic and diverse, posing significant challenges for generalizable detection. While Vision Foundation

arXiv:2502.07645v3 Announce Type: replace Abstract: Behavior cloning (BC) optimizes policies by treating human demonstrations as pointwise action labels. While effective with accurate action labels, t

arXiv:2604.27267v1 Announce Type: cross Abstract: As large language models are integrated into autonomous robotic systems for task planning and control, compromised inputs or unsafe model outputs can

arXiv:2510.05192v2 Announce Type: replace-cross Abstract: When AI agents operating with access to sensitive information encounter a conflict between completing an assigned task and following rules or

arXiv:2601.19768v3 Announce Type: replace Abstract: Large language models (LLMs) are increasingly paired with activation-based monitoring to detect and prevent harmful behaviors that may not be appare

arXiv:2604.28111v1 Announce Type: new Abstract: End-to-end (E2E) autonomous driving presents a promising approach for translating perceptual inputs directly into driving actions. However, prohibitive

Hey @DHLCanadaHelp, your customer service really and truly sucks. You claimed to try to deliver a package, but didn’t actually contact me, didn’t leave a service card, your automated software won’t le

arXiv:2604.26972v1 Announce Type: cross Abstract: This paper studies the computational difficulty of clustering problems that are defined directly on a continuous probability density. Rather than work

arXiv:2604.27147v1 Announce Type: cross Abstract: In generative modeling, we often wish to produce samples that maximize a user-specified reward such as aesthetic quality or alignment with human prefe

This appears to be a casual social media post by AI researcher Gary Marcus discussing competitive dynamics in the AI field, with a conversational reference to Mark Zuckerberg as a notable figure or 't

I have many beefs with Dario and don’t trust him or his hype — but he has certainly eaten OpenAI’s lunch, despite their immense initial lead. Maybe “clown” isn’t the right word here. I think Jensen ag

I just had my first ride with HOVR, a Canadian alternative to Uber. They are cheaper and pay their drivers more. They already have over 3000 drivers in the Greater Toronto Area. You can download HOVR

I should also add that one of the strongest alignment actions that OpenAI did was to name their product chatgpt with gpt 5.5 medium, names so uninspired that nobody could see it as a friend. Unlike Cl

If you think you are going to get alignment out of LLMs you are sadly mistaken. If you live in a society in which people are rolling out LLMs at massive scale, without a robust solution to alignment (

arXiv:2505.13230v3 Announce Type: replace Abstract: Scaling laws in deep learning -- empirical power-law relationships linking model performance to resource growth -- have emerged as simple yet striki

arXiv:2503.01611v3 Announce Type: replace Abstract: Instruction following is a critical ability for Large Language Models to perform downstream tasks. The standard approach to instruction tuning has r

arXiv:2604.28158v1 Announce Type: new Abstract: Existing research infrastructure is fundamentally document-centric, providing citation links between papers but lacking explicit representations of meth

Jensen is one the smartest and most far seeing folks the world. 'If an AI scientist warns people that AI is going to permeate across radiology and radiologists are going to get wiped out, it might see

arXiv:2604.28005v1 Announce Type: new Abstract: Recent advances in large language models (LLMs) have increasingly relied on reinforcement learning (RL) to improve their reasoning capabilities. Three a

arXiv:2604.27713v1 Announce Type: new Abstract: The risks posed by AI features are increasing as they are rapidly integrated into software applications. In response, regulations and standards for safe

arXiv:2604.27448v1 Announce Type: new Abstract: This paper revisits camera pose estimation through the lens of self-supervised pretraining, focusing on inverse-dynamics pretraining as a scalable alter

arXiv:2604.27998v1 Announce Type: cross Abstract: Latent reasoning offers a more efficient alternative to explicit reasoning by compressing intermediate reasoning into continuous representations and s

arXiv:2604.28010v1 Announce Type: cross Abstract: We reframe clinician overrides of clinical AI recommendations as implicit preference data - the same signal structure exploited by reinforcement learn

arXiv:2604.27077v1 Announce Type: cross Abstract: The Normalized Transformer, or nGPT (arXiv:2410.01131) achieves impressive training speedups and does not require weight decay or learning rate warmup

arXiv:2604.27224v1 Announce Type: new Abstract: Quadrupedal loco-manipulation is commonly built on visual perception and proprioception. Yet reliable contact-rich manipulation remains difficult: visio

arXiv:2604.26964v1 Announce Type: cross Abstract: The growing sophistication of contemporary cyber threats necessitates a more effective and adaptive approach to cybersecurity training. Intuitive and

arXiv:2604.27283v1 Announce Type: cross Abstract: Large language model (LLM)-based coding agents increasingly rely on external memory to reuse prior debugging experience, repair traces, and repository

arXiv:2604.27712v1 Announce Type: cross Abstract: Scene-text image captioning requires fusing three information streams -- visual features, OCR-detected text, and linguistic knowledge -- to generate d

arXiv:2604.26960v1 Announce Type: cross Abstract: Transformer-based agentic AI is rapidly being deployed on major platforms to help users shop, watch, and navigate content with less effort. While thes