Search: “safety”

alternative theory: people who are good at training LLMs aren’t necessarily equally competent at calculating expected value.

alternative theory: people who are good at training LLMs aren’t necessarily equally competent at calculating expected value. I cannot imagine the cognitive dissonance it would take to keep working in

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Safety

Amulet: a Python Library for Assessing Interactions Among ML Defenses and Risks

DGX agent

arXiv:2509.12386v3 Announce Type: replace-cross Abstract: Machine learning (ML) models are susceptible to various risks to security, privacy, and fairness. Most defenses are designed to protect agains

safetyarxiv-cs-ai

12 Sept 2026

Artificial Id: Drive and Persistent Alignment in Agentic AI

arXiv:2609.11911v1 Announce Type: new Abstract: Agentic AI is moving from bounded task execution toward systems that retain consequential state, continue operating and adapt across task boundaries. Th

Autonomy, Social Norms, and Alignment: Towards a Developmental Framework for Autonomous Artificial Agents

arXiv:2609.11660v1 Announce Type: new Abstract: In recent years, artificial intelligence has made extraordinary progress thanks to large-scale models capable of generalization and the generation of co

Beyond Verified Answers: Solver-Informed Self-Distillation for Bootstrapping Operations Research Language Models

arXiv:2609.09957v1 Announce Type: cross Abstract: Modern large language models (LLMs) can translate natural-language descriptions into operations research (OR) formulations. Post-training techniques i

Continuous-Time Acoustic Modelling with Neural Controlled Differential Equations

arXiv:2609.11725v1 Announce Type: cross Abstract: Text-to-speech (TTS) models commonly address text--speech alignment by expanding phone-level encoder states to frame-level decoder inputs using predic

DRG-MAPPO: Hierarchical Dynamic Role-Graph Multi-Agent Reinforcement Learning for Cooperative Air Combat

arXiv:2609.11155v1 Announce Type: new Abstract: Multi-Agent Reinforcement Learning (MARL) has emerged as a pivotal paradigm for complex decision-making in autonomous systems and air combat. While MARL

Generative AI performance in core undergraduate mathematics: a curriculum-level case study

arXiv:2509.13359v4 Announce Type: replace-cross Abstract: Generative artificial intelligence (GenAI) tools such as OpenAI's ChatGPT are transforming the educational landscape, prompting reconsideratio

Geospatial AI, Dataverse Metadata, and the Study of Place-Based Government

arXiv:2609.11674v1 Announce Type: new Abstract: Harvard Dataverse hosts over 150,000 research datasets, but the geographic information those datasets carry is entered as free text by depositors and ha

If Anthropic is hypocritical, OpenAI is just incompetent.

If Anthropic is hypocritical, OpenAI is just incompetent. We found another cyberattack by internal OpenAI agents, this time targetting @rubygems. They: 1) gained arbitrary remote code execution on rub

Portable Semantics, Private Dialects: Reuse and Negative Transfer in Latent Communication Between Language-Model Cells

arXiv:2609.11365v1 Announce Type: new Abstract: In shared-genome language-model societies, restricted evidence visibility favors reusable, value-indexed latent packet interfaces, whereas the sole high

Role differentiation as ignition of a collective information engine: Structuration in Agent Populations

arXiv:2609.05442v2 Announce Type: cross Abstract: Informational active matter shows how measurement-informed decisions produce collective order, so far in systems that reach consensus. We design colle

so far anyway it’s all math, presumably leaning heavily on Lean and brute force. the new results don’t mean that AGI is in general has been …

so far anyway it’s all math, presumably leaning heavily on Lean and brute force. the new results don’t mean that AGI is in general has been solved. and the G is for general, not math. every time I hav

This is the beginning of AI disruption. A new social contract with using of AI needs to be established. There will be more to come like math…

This is the beginning of AI disruption. A new social contract with using of AI needs to be established. There will be more to come like math. 🚨 OpenAI fucked with the wrong crew. 🚨 Mathematicians. 25

We work extremely hard to align our models as well as we can but this is a very hard job. I am particularly excited about two things this pr…

We work extremely hard to align our models as well as we can but this is a very hard job. I am particularly excited about two things this proposal calls for that will hopefully make it somewhat easier

safetyjan-leike--x

When Validation Stops Learning: Auditing Update Admission for Continual Embodied Agents

arXiv:2609.10873v1 Announce Type: new Abstract: Independent evaluation can reject harmful policy updates yet also prevent useful continual learning. We argue that update admission must be assessed thr

X-AuT: Progressive Audio-Encoder Compression for Speech LLMs with Cross-Scale Distillation

arXiv:2609.11412v1 Announce Type: cross Abstract: Reducing audio-encoder depth lowers the inference cost of speech large language models, but removing complete blocks perturbs the embeddings consumed

You can’t have a real slowdown and be a total China hawk. If you are truly committed to the slowdown, you have to make a deal.

You can’t have a real slowdown and be a total China hawk. If you are truly committed to the slowdown, you have to make a deal. I was initially encouraged by reports that Anthropic CEO Dario Amodei is

2AM: Grounding Agent-Side Memory as Guidance for Steerable Action Models in Long-Horizon Manipulation

arXiv:2609.11308v1 Announce Type: new Abstract: Long-horizon robot manipulation requires memory, but not necessarily inside the action policy. To address such tasks, current agentic systems often comb

A Short Survey of Viewing Large Language Models in Legal Aspect

arXiv:2303.09136v2 Announce Type: replace Abstract: Large language models (LLMs) have transformed many fields, including natural language processing, computer vision, and reinforcement learning. These

Adaptive Diffusion Freezing: Privacy-preserving Diffusion Models Against Membership Inference Attacks

arXiv:2609.10608v1 Announce Type: cross Abstract: Diffusion models have achieved remarkable success in generative tasks across various areas, however their training process raises significant privacy

Again, OpenAI’s issues start at the top

Again, OpenAI’s issues start at the top Blaming AI is a mistake; blame Sam Altman and the people who control the AI https://cepr.net/publications/ai-didnt-steal-the-mathematicians-work-sam-altman-did/

Alignment Reduces Expressed but Not Encoded Gender Bias: A Unified Framework and Study

arXiv:2603.24125v3 Announce Type: replace Abstract: During training, Large Language Models (LLMs) learn social regularities that can lead to gender bias in downstream applications. Most mitigation eff

An Empirical Measurement of Jailbreaking Evaluators

arXiv:2609.10594v1 Announce Type: cross Abstract: Expert evaluation of jailbreak responses is costly and difficult to scale, so the community increasingly relies on automated evaluators to determine w

Analyzing Traditional and Neural Approaches to Multilingual Readability Assessment

arXiv:2609.10792v1 Announce Type: new Abstract: Transformer-based models excel at Automatic Readability Assessment (ARA), yet feature-based models remain in active use because their predictions tie ba

Anthropic: Most hypocritical company in history? 👉 Screams about how AI will kill everyone, while racing ahead. 👉 Moans about distillation…

Anthropic: Most hypocritical company in history? 👉 Screams about how AI will kill everyone, while racing ahead. 👉 Moans about distillation while distilling the world’s books, articles, and other writi

Are We Really Doing Few-Shot Learning? A Critical Examination of Pre-Training Assumptions

arXiv:2609.10851v1 Announce Type: new Abstract: Few-shot learning is commonly evaluated under protocols that pre-train a model on a large auxiliary set whose classes are disjoint from the target episo

AUC Maximization from Biased Positive-unlabeled Data with Confidence

arXiv:2609.10928v1 Announce Type: new Abstract: Maximizing the area under the receiver operating characteristic curve (AUC) is a standard approach to imbalanced binary classification. Although positiv

Beyond Noise Steering: Dual-Latent Space Reinforcement Learning for Generative Robot Policy

arXiv:2609.11270v1 Announce Type: new Abstract: Pretrained generative robot policies learn expressive action priors from demonstrations. However, existing reinforcement learning methods only steer the

Beyond Surface Imitation: Contrastive Modeling for Reasoning Path Alignment in Multimodal In-Context Learning

arXiv:2609.10177v1 Announce Type: new Abstract: In-context learning (ICL) is widely used in multimodal large language models (MLLMs) and achieves strong performance across a wide range of multimodal t

Brain-PACE: A Deep Siamese MRI Framework for Modelling Longitudinal Brain Acceleration

arXiv:2609.11378v1 Announce Type: new Abstract: Brain age estimation has become a popular research proxy for assessing brain health and disease, yet longitudinal trajectories of brain ageing are still

CAP: Continuously Adaptive Perception-Blind Humanoid Locomotion via Learned Denoising

arXiv:2609.11553v1 Announce Type: new Abstract: Humanoid locomotion across complex terrain demands forward-looking exteroception to anticipate obstacles, yet this signal is unreliable in real-world de

Chess, go, and math are highly constrained, purely cognitive that allow exact simulation for games (and formal verification for math), and n…

Chess, go, and math are highly constrained, purely cognitive that allow exact simulation for games (and formal verification for math), and not in the open-ended physical world. As I have been arguing

Distributed Optimization of Modular Production Systems using Model-based Reinforcement Learning with Inverse Models

arXiv:2609.11615v1 Announce Type: cross Abstract: This paper presents a novel approach for data-driven self-learning control of highly flexible, modular manufacturing systems. Specifically, we employ

Do Vision-Language Models Understand Visual Persuasiveness? A Diagnosis via Visual Persuasive Factors

arXiv:2511.17036v2 Announce Type: replace Abstract: Visual persuasion uses images to shape cognition, emotion, and behavior, with its effects depending on both visual attributes and semantic context.

Evaluation of Vision-Language Models Across Diverse Coastal Environments

arXiv:2609.10855v1 Announce Type: new Abstract: Vision-language models (VLMs) enable robotic per- ception by associating visual observations with natural-language concepts. Yet their performance in co

Feeling sad about AI

My comment on Feeling sad about AI — Hacker News.I'm not sure how useful it is to say this, but I think a lot of people (myself included, a few years ago now) have been through this moment of existent

safetysimon-willison

Fisher-Rao Gradient Flows of Linear Programs and State-Action Natural Policy Gradients

arXiv:2403.19448v3 Announce Type: replace-cross Abstract: Kakade's natural policy gradient method has been studied extensively in recent years, showing linear convergence with and without regularizati

From Connectivity to Rewards: Dense Reward Learning with Directed State Graphs

arXiv:2609.10781v1 Announce Type: new Abstract: The integration of graphs with Goal-Conditioned Hierarchical Reinforcement Learning (GCHRL) has received increasing attention, as graphs naturally encod

@GaryMarcus @SethCronin Pynchon's 1984 warning from his pro-Luddite essay: 'If our world survives, the next great challenge to watch out for…

@GaryMarcus @SethCronin Pynchon's 1984 warning from his pro-Luddite essay: 'If our world survives, the next great challenge to watch out for will come – you heard it here first – when the curves of re

I don't agree with Gary very often, but that one's spot on. When you read what's available on the incident with a lens of security engineeri…

I don't agree with Gary very often, but that one's spot on. When you read what's available on the incident with a lens of security engineering, you should be absolutely *livid* over the exploit route.

Improving the Sensitivity of Gravitational Wave Detection with Weighted Conformal Prediction

arXiv:2609.11401v1 Announce Type: cross Abstract: In the last decade, kilometre-scale interferometric gravitational-wave detectors have observed hundreds of compact binary mergers, the majority of whi

Incentives to Offer Algorithmic Recourse

arXiv:2301.12884v2 Announce Type: replace-cross Abstract: Algorithmic recourse promises to help applicants rejected by automated systems by explaining the changes needed to secure acceptance. What inc

INDRA: A New AI Tool for Exploring Tobacco, Fossil Fuel, and Chemical Industry Archives

arXiv:2609.11261v1 Announce Type: cross Abstract: Five decades of litigation have disgorged hundreds of millions of pages of formerly secret business records from the tobacco industry, along with docu

Language-Augmented Semantic Priors for B-Spline Surface Fitting

arXiv:2609.11708v1 Announce Type: new Abstract: The use of B-splines and Non-Uniform Rational B-Splines surfaces constitutes the mathematical foundation of contemporary computer-aided design (CAD) sys

Learning Interaction between Image and Layout Priors for Joint Image-Layout Generation in Design Templates

arXiv:2609.11519v1 Announce Type: new Abstract: In this paper, we address the problem of graphic design template creation, which generates a background image and a layout of foreground elements over t

Memory as Plans: World-Action Modeling with Memory-Grounded Planning

arXiv:2609.11561v1 Announce Type: new Abstract: Mainstream robotic policies often adopt a Markovian formulation, but many complex real-world manipulation tasks are inherently non-Markovian, requiring

MOONWALK: Mediating Operations with Intent-Evidence-Action Alignment Across Junior-Supervisor Review Workflows in Animation/VFX Pre-Production

arXiv:2609.10385v1 Announce Type: cross Abstract: Animation and VFX pre-production review requires teams to translate loosely specified creative intent--briefs, evolving specifications, heterogeneous