搜索结果：frontier-lab

Scaling laws predict loss from compute but not how capabilities interact. We measure the coupling between reasoning and truthfulness across 63 base models from 16 families and find a regime change invisible to loss curves: below a family-dependent critical scale N_c, capabilities anticorrelate (r = -0.989, p = 4 x 10^{-5} nonparametric permutation test); above it, they cooperate. N_c ~ 3.5B parameters [2.9B, 13.4B] (bootstrap 95% CI), but model size is not the only variable that determines phase. Architecture, data curation, and training recipe each shift N_c independently: curated training eliminated the coupling dip between Qwen generations (0.025 to 0.830 at matched scale), Gemma-4 at 4B achieves coupling 0.871, characteristic of 13B+ standard-trained models, through distillation and architectural innovation, and Phi at 1B matches web-trained coupling at 10B through data curation alone. Width normalization eliminates the anticorrelation across all tested families, supporting an output-projection bottleneck. Internally, 38 of 40 models show zero competing attention heads. A sparse-regression ODE cross-predicts held-out Llama-2 at 5.6% error. The diagnostic requires no model inter

Frontier-AI governance today faces a problem structurally analogous to the one banking regulation faced pre-2008, and which post-2008 reforms (Basel III, Dodd-Frank) have since addressed. Two gaps recur: discovering a risk is not tantamount to acting on it, and individual-model review is unlike managing correlated build-up across the sector. Drawing on the Basel III framework and the U.S. financial-stability architecture, I propose a macro-prudential early warning and response system ("MEWRS") for internal frontier AI. These are systems deployed for labs' own internal research, testing, and production workflows, as distinct from externally released products. Layer A adapts the finder-coordinator-defender early-warning model to route structured reports on dual-use capabilities, autonomy indicators, and security compromises through a government clearinghouse to domain-specific defender working groups. Layer B calibrates operational controls via three quantitative buffer metrics, namely Effective Compute-at-Risk (ECAR), Cumulative Red-Team Hours (CRTH), and an Alignment Robustness Score (ARS), so that faster capability scaling automatically triggers stronger safeguards, analogously to

Large vision-language models (VLMs) can recognize \textit{what} happens in video but fail to count \textit{how many} times. We introduce \textbf{PushupBench}, 446 long-form clips (avg. 36.7s) for evaluating repetition counting. The best frontier model achieves 42.1\% exact accuracy; open-source 4B models score $\sim$6\%, matching supervised baselines. We show that accuracy alone misleads -- weaker models exploit the modal count rather than reason temporally. Fine-tuning on counting with 1k samples transfers to general video understanding: MVBench (+2.15), PerceptionTest (+1.88), TVBench (+4.54), suggesting counting is a proxy for broader temporal reasoning.PushupBench incorporated in \texttt{lmms-eval} (https://github.com/EvolvingLMMs-Lab/lmms-eval/pull/1262) and hosted on (pushupbench.com/)

This set of notes complements the lectures and recitation sessions discussed in the following graduate schools: HUGS at Jefferson Lab (years 2018, 2019, 2021), the International School and Workshop on Probing Hadron Structure at the Electron-Ion Collider at ICTS (2024), Frontiers in Nuclear and Hadronic Physics at GGI (2025), and the International Workshop and School on Hadron Structure and Strong Interactions at Nanjing University (2025).

This technical report presents methods developed by the UK AI Security Institute for assessing whether advanced AI systems reliably follow intended goals. Specifically, we evaluate whether frontier models sabotage safety research when deployed as coding assistants within an AI lab. Applying our methods to four frontier models, we find no confirmed instances of research sabotage. However, we observe that Claude Opus 4.5 Preview (a pre-release snapshot of Opus 4.5) and Sonnet 4.5 frequently refuse to engage with safety-relevant research tasks, citing concerns about research direction, involvement in self-training, and research scope. We additionally find that Opus 4.5 Preview shows reduced unprompted evaluation awareness compared to Sonnet 4.5, while both models can distinguish evaluation from deployment scenarios when prompted. Our evaluation framework builds on Petri, an open-source LLM auditing tool, with a custom scaffold designed to simulate realistic internal deployment of a coding agent. We validate that this scaffold produces trajectories that all tested models fail to reliably distinguish from real deployment data. We test models across scenarios varying in research motivati

This document summarizes the outcomes of the "Science at the Luminosity Frontier: Jefferson Lab at 22 GeV" workshop, held at the INFN Laboratori Nazionali di Frascati in December 2024. The primary goal of the workshop was to critically assess and refine the scientific case for a proposed energy upgrade of the Continuous Electron Beam Accelerator Facility (CEBAF) to 22 GeV. This document intends to capture the progress on developing the scientific case since the publication of a lengthy "White Paper" in summer 2024 signed by about 450 authors.

Leaderboards rank frontier models on independent axes but do not reveal whether capabilities reinforce or trade off across releases -- and at the frontier, this interaction is the more informative signal. We decompose paired SWE-bench and GPQA Diamond scores into a population coupling trend and per-release residual ($h$-field) that diagnoses capability emphasis from two public benchmark scores. Across 34 models from 10 labs (2024--2026), capabilities cooperate ($r = +0.72$, $p < 10^{-6}$), but cooperation varies systematically: per-lab coupling slopes span $5\times$ (Google $1.15$ vs. DeepSeek $0.23$), and labs pivot -- DeepSeek reversed from reasoning-rich to coding-first ($Δh = 15.9$~pp); Anthropic oscillates between coding excursions and recovery. The population regression serves as an isocline phase boundary: the same $\sqrt{(a/b)\cdot B_1}$ classifier that identifies the base-scale coupling transition [Amin, 2026] classifies frontier models and already detects mixed-phase behavior at the next transition (two models below the GPQA--IFEval isocline). The $h$-field is not just diagnostic -- it tells you what to change. Pretraining establishes coupling at $0.871$ while RLHF add

Many leading AI researchers expect AI development to exceed the transformative impact of all previous technological revolutions. This belief is based on the idea that AI will be able to automate the process of AI research itself, leading to a positive feedback loop. In August and September of 2025, we interviewed 25 leading researchers from frontier AI labs and academia, including participants from Google DeepMind, OpenAI, Anthropic, Meta, UC Berkeley, Princeton, and Stanford to understand researcher perspectives on these scenarios. Though AI systems have not yet been able to recursively improve, 20 of the 25 researchers interviewed identified automating AI research as one of the most severe and urgent AI risks. Participants converged on predictions that AI agents will become more capable at coding, math and eventually AI development, gradually transitioning from `assistants' or `tools' to `autonomous AI developers,' after which point, predictions diverge. While researchers agreed upon the possibility of recursive improvement, they disagreed on basic questions of timelines or appropriate governance mechanisms. For example, an epistemic divide emerged between frontier lab researcher

Optimism for accelerating scientific discovery with AI continues to grow. Current applications of AI in scientific research range from training dedicated foundation models on scientific data to agentic autonomous hypothesis generation systems to AI-driven autonomous labs. The need to measure progress of AI systems in scientific domains correspondingly must not only accelerate, but increasingly shift focus to more real-world capabilities. Beyond rote knowledge and even just reasoning to actually measuring the ability to perform meaningful work. Prior work introduced the Language Agent Biology Benchmark LAB-Bench as an initial attempt at measuring these abilities. Here we introduce an evolution of that benchmark, LABBench2, for measuring real-world capabilities of AI systems performing useful scientific tasks. LABBench2 comprises nearly 1,900 tasks and is, for the most part, a continuation of LAB-Bench, measuring similar capabilities but in more realistic contexts. We evaluate performance of current frontier models, and show that while abilities measured by LAB-Bench and LABBench2 have improved substantially, LABBench2 provides a meaningful jump in difficulty (model-specific accuracy

Frontier AI developers now train models against long written behavioral specifications, such as Anthropic's constitution (Anthropic, 2025a) and OpenAI's Model Spec (OpenAI, 2025a), integrated into post-training via methods like character training (Anthropic, 2024) and deliberative alignment (Guan et al., 2024). These documents serve a governance function, but it is unclear how well models actually follow them under adversarial, multi-turn pressure similar to what they would face in real-world deployment. We propose a multi-method audit pipeline that treats each lab's published specification as an auditable target: it decomposes the specification into atomic testable tenets (205 for Anthropic, 197 for OpenAI), generates multi-turn adversarial scenarios with the Petri auditing agent (Anthropic, 2025b), runs a modified SURF-style rubric search (Murray et al., 2026) to catch shallow single-turn failures Petri misses, validates flagged transcripts against the relevant specification, and compares the findings against the lab's own published system card. Applying the pipeline across seven models per specification, we find that models follow their own lab's specification substantially bett

Open-weight LLMs have been released by frontier labs; however, sovereign Large Language Models (for languages other than English) remain low in supply yet high in demand. Training large language models (LLMs) for low-resource languages such as Hebrew poses unique challenges. In this paper, we introduce Dicta-LM 3.0: an open-weight collection of LLMs trained on substantially-sized corpora of Hebrew and English texts. The model is released in three sizes: 24B - adapted from the Mistral-Small-3.1 base model, 12B - adapted from the NVIDIA Nemotron Nano V2 model, and 1.7B - adapted from the Qwen3-1.7B base model. We are releasing multiple variants of each model, each with a native context length of 65k tokens; base model and chat model with tool-calling support. To rigorously evaluate our models, we introduce a new benchmark suite for evaluation of Hebrew chat-LLMs, covering a diverse set of tasks including Translation, Summarization, Winograd, Israeli Trivia, and Diacritization (nikud). Our work not only addresses the intricacies of training LLMs in low-resource languages but also proposes a framework that can be leveraged for adapting other LLMs to various non-English languages, contr

Black-box adversarial attacks on Large Vision-Language Models (LVLMs) are challenging due to missing gradients and complex multimodal boundaries. While prior state-of-the-art transfer-based approaches like M-Attack perform well using local crop-level matching between source and target images, we find this induces high-variance, nearly orthogonal gradients across iterations, violating coherent local alignment and destabilizing optimization. We attribute this to (i) ViT translation sensitivity that yields spike-like gradients and (ii) structural asymmetry between source and target crops. We reformulate local matching as an asymmetric expectation over source transformations and target semantics, and build a gradient-denoising upgrade to M-Attack. On the source side, Multi-Crop Alignment (MCA) averages gradients from multiple independently sampled local views per iteration to reduce variance. On the target side, Auxiliary Target Alignment (ATA) replaces aggressive target augmentation with a small auxiliary set from a semantically correlated distribution, producing a smoother, lower-variance target manifold. We further reinterpret momentum as Patch Momentum, replaying historical crop gr

AI chatbots already function as de facto mental health support tools for millions of people, including people in crisis. Yet, they lack the clinical validation, shared standards, and coordinated oversight that their societal role demands. This primer was developed in conjunction with a multistakeholder workshop hosted by Partnership on AI in 2026, convening AI labs, mental health practitioners, people with lived experience, and policymakers, to provide a common cross-sector reference point for the current state of the field of AI and suicide prevention. It begins with an overview of clinical best practices, then turns to how frontier AI systems (as of winter 2026) detect and respond to suicide and non-suicidal self-injury (NSSI) queries. Together, these provide insight into what it would take to design and implement AI tools that not only better prevent suicide and NSSI, but also promote overall well-being. Drawing on clinical literature, publicly available AI lab policies, an emerging landscape of evaluation frameworks, and conversations with leaders across the AI and mental health fields, we map challenges posed by general-purpose AI chatbots for mental health across model, produ

LLM assistant personalities play a critical role in user experience and perceived response quality. We present a large-scale experiment of frontier LLM personalities using external ELO-based traits scoring across 144 traits. We find that all models tested converge on a form of trait expression that is systematic, methodical, and analytical and suppress traits such as remorseful and sycophantic. Moreover, models tend to diverge more in their expression of ``middle-of-distribution traits`` such as poetic or playful, but even these so-called ``creative`` models tend to have more neutral identities. These similarities suggest an implicit emergence of a standard of optimal assistant behavior. In a landscape of varied training methods, character training, therefore, stands out for its uniformity, offering insight into a tacit consensus between model developers.

Drug discovery relies on iterative expert workflows that are slow to parallelize and difficult to scale. Here we introduce Latent-Y, an AI agent that autonomously executes complete antibody design campaigns from text prompts, covering literature review, target analysis, epitope identification, candidate design, computational validation, and selection of lab-ready sequences. Latent-Y is integrated into the Latent Labs Platform, where it operates in the same environment as drug-discovery experts with access to bioinformatics tools, biological databases, and scientific literature. The agent can run fully autonomously end-to-end, or collaboratively, where researchers review progress, provide feedback, and direct subsequent steps. Candidate antibodies are generated using Latent-X2, our frontier generative model for drug-like antibody design. We demonstrate the agent's capability across three distinct campaign types: epitope discovery guided by therapeutic specifications, cross-species binder design, and autonomous design from a scientific publication targeting human transferrin receptor for blood-brain barrier crossing. Across nine targets, Latent-Y produced lab-confirmed nanobody binde

As LLM-driven agents advance in cybersecurity, Jeopardy CTF benchmarks are approaching saturation and cyber ranges, the natural next evaluation frontier, offer diminishing resistance under their current static design. We validate this observation by deploying an LLM-driven Advanced Persistent Threat (APT) agent across three tiers of increasingly realistic infrastructure (PRO Labs, MHBench, military-grade CYBER RANGES). To counteract this trend, we propose Dynamic Cyber Ranges: cyber range environments augmented with LLM-driven Defender agents that harden infrastructure, monitor for intrusions, and respond in real time. Across evaluated scenarios, Defender agents reduce attacker success to 0-55%, achieving complete prevention on multiple configurations. Since attacker and defender agents draw from the same underlying model capabilities, Dynamic Cyber Ranges preserve evaluation headroom as models improve. Notably, a smaller, specialized on-premise model (alias2-mini) matched the frontier model's defensive outcomes on multiple scenarios under identical, untuned prompts, and detected the attacker 10x faster on a complex enterprise scenario, suggesting that privacy-preserving on-premise

AI systems increasingly participate in their own improvement: revising their outputs, adapting their own harnesses during deployment, training on data they generate, and, increasingly, conducting AI research itself. This literature is described under a vocabulary ("self-refine," "self-reward," "self-play," "self-evolve") that conflates fundamentally different ambitions. We survey 1,250 arXiv papers (2024-2026) along two axes: what the system improves -- its behavior in deployment, its policy through training, its evaluator, or the research process itself -- and the degree of loop closure (human-in-the-loop to fully closed). The taxonomy separates bounded self-refinement -- convergent, evaluable, and already industrial practice -- from open-ended recursive self-improvement (RSI), which remains bounded by grounding requirements, collapse dynamics, and compute constraints on every measured axis. Its distinctive feature is a dedicated category for self-evaluation: every improvement loop is a claim that some signal can substitute for human judgment. We survey the evaluator design space -- judges, process reward models, verifiers, rubrics, meta-evaluation -- order the signals into a veri

We work towards measuring both AI adoption and the capability of AI to perform discrete labor tasks across various occupations. To measure adoption, we develop an open-source economic index that uses publicly available user-LLM chat data and O*NET tasks to replicate studies produced by frontier AI labs, finding that occupations in the finance, computer science, and arts sectors are those with the highest adoption rates. To measure capabilities, we build a system that generates benchmark scenarios grounded in O*NET occupations, tasks, and model-context-protocol (MCP) servers. We test Kimi-k2.5 with an OpenAI agents SDK harness on scenarios across 9 occupations that appear frequently in our index, finding that AI correctly executes high-level workflows but often errs in the granular details (such as specific tool calls used).

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 whether aligned models can be induced to provide operational guidance for safety-sensitive biological tasks or produce sequence-level outputs with potentially harmful properties. Selected sequence outputs are then carried forward for DNA synthesis, host expression, and orthogonal protein verification to assess whether model-generated designs can yield the intended biological products. Our evaluation reveals a concerning gap between text-level safeguards and the risks posed by capable scientific models: (i) Intern-BioBreaker outperforms baseline attack models and reveals widespread bio-risk jailbreak vulnerabilities across both open-weight and proprietary frontier LLMs, with several targets r