Safety & alignment

The growing adoption of artificial intelligence in judicial applications exposes a critical limitation of neural network-based legal AI systems: their decision-making processes are inherently opaque, which undermines the reliability of their outcomes, especially in cases involving personal liberty. Existing interpretability methods primarily rely on instance-based analysis and local input-output attributions; however, they are unable to capture the models’ global decision-making mechanisms, whic

Direct feedback alignment (DFA) trains hidden layers with fixed random projections of the output error, avoiding the transposed-weight backward pass of backpropagation (BP). We study a failure mode of DFA training that is distinct from feedback quality: the local weight update is calculated by an outer product, so anisotropy can enter through either its presynaptic-activity factor or its local-error factor. Our analyses with controlled synthetic regimes isolate the first failure mode and show an

Discourse relations provide document structure, critical to language understanding and enabling language model performance and ethicality. In this work, we investigate how instruction-tuned Transformer models (LLaMA and Mistral) encode discourse relations in English, with a particular focus on the contrasting relations of causation and antithesis. Framing the task as a next-token prediction task and applying a suite of interpretability techniques to test model internals, our findings show that c

A graph foundation model generalizes across graph domains by mapping every input into one shared representation before any task reasoning. We call this map the alignment layer, the component that separates a graph foundation model from a graph neural network, and we show it is a distinct attack surface that prior work has not studied. We attack it at inference time, with no access to training, on six public models spanning spectral tokenizers, text embedding spaces, and a discrete codebook. A di

AI alignment is essential for the safe deployment of advanced AI systems. Given that values and preferences change over time, culture, social roles, and context, we need to develop a better understanding of the possible long-term consequences of AI alignment, in particular considering the likely ubiquitous future use of personalized AI assistants. We introduce a flexible and extensible mathematical modelling framework, rooted in social physics, aimed at answering macro-level questions regarding

Arvind Raman, the director of National Institute of Standards and Technology, will serve as acting director of CAISI, according to a spokesperson

The digital substrate of states -- data, algorithms, infrastructure, platforms, applications -- is being governed without adequate conceptual foundations. The ability and legitimacy required to govern this substrate, and to govern with it, are simultaneously misaligned, contested, and structurally absent. We introduce digital statecraft as the organising concept for this emerging field, arguing that 'digital' reconstitutes the statecraft question rather than merely extending its domain. The conc

Text-to-image diffusion transformers (DiTs) jointly process text and image tokens, yet their internal computation during denoising remains poorly understood. We introduce a causal interpretability framework for modern large-scale DiTs that combines attention decomposition with targeted interventions across token spans, heads, and layers. Using it to separate prompt-content tokens from structural template tokens, we find that the structural tokens carry little prompt-specific information at the e

The rapid improvement of AI systems has intensified debate about humanity's economic, political, social, and existential future. As AI reshapes expectations about what lies ahead, policy choices and institutional responses will play a crucial role in determining who benefits, who bears the costs, and whether the most serious risks can be mitigated. Africa remains relatively overlooked in these discussions, partly because it is largely a consumer rather than a producer of frontier AI systems, and

Dr. Chris Fall, the director of the Commerce Department's Center for AI Standards and Innovation (CAISI) is resigning after just three months in the role. The exit of Fall, who was appointed to head the AI safety organization in April, comes amid a fluctuation in the Trump administration's AI policy. Staff at CAISI are directly...

A new, gonzo style of fraud could be an existential threat to prediction markets.

Large language models (LLMs) serving as planners in tool-using autonomous agents introduce dynamic reliability risks in multi-turn execution. While single-turn safety mechanisms are relatively mature, extended interactions reveal structural vulnerabilities where initial alignment degrades over time. This paper empirically characterizes two observed failure modes across multiple state-of-the-art LLMs: Safety Drift, the gradual erosion of declared safety intent leading to constraint-violating acti

Center for AI Standards and Innovation Director Chris Fall played a key role in the White House’s voluntary review of cutting-edge models.

Modern LLMs are alarmingly susceptible to surprisingly simple immaterial changes of input prompts: a casual hint, an incorrectly labeled few-shot example, or a fake prior assistant turn often flips an originally correct answer. We study where this susceptibility, spanning sycophancy and related cue-induced biases, lives inside the model. Across five model families and seven BCT bias types, we extract a per-bias direction from hidden states and triangulate it through three measures: probing, leav

Frontier large language models (LLMs) are increasingly integrated into scientific workflows, yet their growing biological capabilities may outpace current safeguards. To assess the biological risks of frontier models, we develop Intern-BioBreaker, a specialized bio-red-teaming model, together with an integrated computational-to-physical framework that couples model-level stress testing with wet-lab validation. Within this framework, Intern-BioBreaker generates targeted jailbreak prompts to test

Scaled dot product attention conflates directional alignment and vector magnitude, limiting its effectiveness as a similarity metric in Transformer models. We introduce L1 augmented attention, a simple and computationally parallelizable modification that subtracts a learned, head specific L1 distance between queries and keys from the dot product score. This hybrid similarity captures complementary geometric information. Dot product rewards directional alignment, while L1 penalizes coordinate dev

Can large language models with substantially different parameter spaces be merged by direct weighted averaging, without training or semantic alignment? Existing heterogeneous fusion methods typically introduce distillation, adapters, learned latent spaces, routing, or feature alignment, leaving open whether a simpler recipe can work for genuinely different billion-parameter checkpoints. We revisit this counterintuitive question through training-free dimensional adaptation followed by ratio-contr

The prevailing inference framework for diffusion models formulates generation fundamentally as a problem of numerical integration. This perspective casts the model as an exact estimator, neglecting the inherent statistical uncertainty of the denoising process. In this work, we propose Forward-Process Aligned Diffusion prediction (\textbf{DiFA}), a training-free framework that reframes inference-time data prediction refinement as a sequential state estimation problem. Rather than reusing past out

Generative artificial intelligence is rapidly transforming materials design by enabling de novo exploration of immense chemical spaces. Yet a large proportion of AI-generated compositions remain implausible, violating established chemical principles, which limits the reliability and interpretability of generative materials design. Here, we introduce a chemical validity operator that recasts heuristic chemical rules as a configurable algorithmic prior for evaluating and guiding generative materia

The assessment of jailbreak attacks against large language models currently suffers from inconsistent evaluation criteria and methods, leading to unreliable estimates of attack success rates. We propose JailMeter, an evidence-based evaluation framework designed to more faithfully measure jailbreak effectiveness. Inspired by the Information Bottleneck theory, JailMeter applies dual-feedback optimization to filter jailbreak noise from model responses while preserving content relevant to the origin

Why This Matters The Office of Personnel Management (OPM) is the federal government's central agency for human resources and workforce management. Since December 2024, in alignment with presidential directives on the closure of certain offices and the reduction of the size of the federal workforce, OPM has undergone significant workforce and organizational structure changes. These actions have reduced institutional knowledge and operational capacity at the agency. OPM’s Headcount Decreased by 35

Accurate 3D--2D liver registration, which aligns preoperative 3D models to partial, view-dependent intraoperative surface observations, is critical for AR-guided laparoscopic surgery but remains challenging due to severe occlusion, limited visibility, and the lack of 3D ground-truth supervision. Existing landmark-free approaches perform partial-to-complete geometric alignment, yet robust self-supervision under extreme partial visibility remains difficult. We propose Vis2Reg, a visibility-aware r

Vision-Language Navigation (VLN) requires an embodied agent to interpret a natural-language instruction and predict actions from temporally ordered visual observations. Adapting a multimodal large language model to VLN requires visual-language alignment, compact temporal inputs, action-space grounding, and stable training on the target hardware. This technical report presents PGN (Pangu Navigator), an offline VLN action-prediction system built on OpenPangu-7B. Training proceeds in two stages. Fi

Text-to-Image (T2I) generative models have achieved remarkable progress in synthesizing high-quality visual content, yet they remain vulnerable to adversarial misuse, particularly in generating Not-Safe-For-Work (NSFW) images. Most existing jailbreak attacks primarily rely on heuristic prompt engineering or black-box optimization, treating model feedback as a binary signal (success or failure). This coarse-grained paradigm overlooks the rich information embedded in diverse failure modes, such as

Within Explainable Artificial Intelligence, mechanistic interpretability uses Sparse Autoencoders (SAEs) to extract more interpretable features from neural representations. However, assessing their monosemanticity, and thus explanation quality, remains challenging. Existing metrics require external concept labels or depend on pretrained embedding models, making them sensitive to encoder's geometry. We introduce the Tversky Monosemanticity Score (TMS), a label-free metric that operationalizes mon

OpenAI shares lessons from deploying long-running AI models, highlighting new safety risks, observed failures, and improved safeguards through iterative deployment.

Contrastive learning is increasingly moving toward settings with three or more modalities instead of image-text pairs. Yet, extending models from pairwise to higher-order multimodal alignment can introduce optimization and representation challenges. We identify encoder Jacobian conditioning as a key factor in trimodal contrastive learning: poorly conditioned encoders exhibit collapsing or amplified singular-value spectra, leading to exploding Jacobian condition numbers and degraded multimodal al

Unsupervised Graph Domain Adaptation (UGDA) aims to facilitate knowledge transfer from a labeled source graph to an unlabeled target graph by mitigating cross-domain distribution shifts. Existing methods primarily focus on node-level feature alignment in latent spaces, relying on the implicit assumption that all source nodes contribute positively to the alignment. However, this assumption often fails because a node's semantic information is intrinsically coupled with its topological graph struct

Mixed Reality (MR) is increasingly being used in exhibition settings to bring digital artworks into relation with the physical environment. However, existing MR exhibition systems are often confined to prototypes or case-specific deployments, offering limited guidance for large-scale practical implementation. To address this gap, this paper presents a practical pipeline for designing and deploying large-scale MR art exhibitions, treating spatial alignment not only as a technical mechanism but al

Multimodal large language models (MLLMs) still struggle with spatial reasoning that requires perspective transformation. In particular, they often rely on camera-centric cues rather than reasoning from the reference object's viewpoint, leading to systematic errors in non-camera reference settings. In this paper, we first analyze this failure mode and show that object orientation is a key factor underlying such camera-centric shortcut behavior. To address this issue, we propose OrientSAM, an orie

Modern video transformers typically ignore principles from primate vision and are rarely evaluated against neural data, limiting their biological interpretability. We introduce a sparse winner-takes-all token selection module that replaces dense self-attention to improve efficiency and approximate competitive routing observed in biological visual circuits. We further propose a neuro-inspired split-and-fuse video transformer which uses two complementary pathways: a high-resolution, low-frame-rate

arXiv:2607.02672v2 Announce Type: replace-cross Abstract: Local pairwise comparisons are a standard tool for learning how people want decision rules to work, e.g., in participatory design or alignment. However, their use builds in two strong assumptions: that local comparisons are sufficient evidence about how a person wants an automated decision rule to behave, and that people can always answer those comparisons decisively. We investigate how these assumptions may be compromised under internal

Long video summarization presents significant challenges for multimodal large language models (MLLMs), particularly in maintaining temporal fidelity over extended durations and producing summaries that are both semantically and temporally grounded. We introduce LVSum, a human-annotated benchmark for evaluating long-form video summarization with fine-grained temporal alignment. LVSum comprises 72 diverse videos spanning 13 domains with an average duration of 16 minutes, each annotated with up to

The prevailing inference framework for diffusion models formulates generation fundamentally as a problem of numerical integration. This perspective casts the model as an exact estimator, neglecting the inherent statistical uncertainty of the denoising process. In this work, we propose Forward-Process Aligned Diffusion prediction (DiFA), a training-free framework that reframes inference-time data prediction refinement as a sequential state estimation problem. Rather than reusing past outputs sole

Understanding physical human-robot and human-human interactions is a challenging yet emerging topic in 3D vision. While most existing methods rely on skeleton sequences--effective in low-light and privacy-sensitive environment--they face two major challenges: 1) learning and effectively exploiting interaction cues from skeletal data, and 2) compensating for the lack of visual information absent in skeletons alone. To address these challenges, we propose skeletal token alignment and rearrangement

Given a binary-labeled linearly separable dataset, and the objective is to compute the maximum-margin separating hyperplane, also known as the hard-margin Support Vector Machine (SVM) classifier. This paper investigates whether, if given an initial separating hyperplane, can it be exploited to reach this unique optimum more efficiently. We present a geometric approach that gradually improves the alignment of the hyperplane, starting from an initial separating hyperplane, while preserving separat

Recently, text-to-video (T2V) models have been widely deployed, sparking growing concerns over their robustness against jailbreak attacks. Existing jailbreak methods, mostly adapted from text-to-image attacks, suffer notable drawbacks when applied to T2V systems. They fail to fully leverage temporal consistency, an inherent characteristic of video generation. Besides, these methods demand heavy video query optimization, which is infeasible in practical black-box scenarios. Their adversarial prom

Large language model (LLM) post-training is essential for improving reasoning, adaptation, and alignment. Existing methods mainly follow two paradigms: reinforcement learning (RL) and on-policy distillation (OPD). However, RL relies on coarse-grained outcome supervision, resulting in difficult credit assignment and limited capability to acquire new knowledge. OPD, meanwhile, unconditionally matches teacher logits through KL divergence, which creates a dilemma: similar teachers provide little new

Human-like private chat requires more than fluent response generation: a system must preserve persona, relationship, memory, bounded knowledge, medium-specific timing, and a coherent multi-turn arc. We present AnthroDial, a closed-loop framework that formulates anthropomorphic dialogue as a joint problem of system architecture, executable evaluation, and diagnostic alignment. It combines (1) a role-conditioned scheduled dialogue runtime with persona and scenario cards, long-term memory, virtual

Test-time scaling empowers Large Reasoning Models (LRMs) to tackle complex tasks via extensive Chain-of-Thought (CoT). However, this often induces the "overthinking" paradox, where redundant reasoning increases computational overhead without guaranteeing accuracy. Existing test-time efficiency optimization methods primarily fall into two categories: information-theoretic approaches, which are prone to "deceptive convergence" where low uncertainty masks hallucinations, and latent representation a