Claude cracks HAWK, a video model drives robots, Gemma reads materials physics

Sources

Discovering cryptographic weaknesses with Claude anthropic.com

Discovering cryptographic weaknesses with Claude simonwillison.net

Discovering cryptographic weaknesses with Claude The best part of this article (here’s the repo ) about how Anthropic researchers used Claude Mythos to find mathematical flaws in both HAWK and a weaker version of AES (“neither of these results has a practical impact on today’s computer systems”) is the prompts that they shared, spelling mistakes included: the models tend to think it is impossible to solve so they don’t try they need a good amount of prompting. why not do aes-128 r7? the whole p…

Masked Visual Actions for Unified World Modeling huggingface.co

Video models absorb rich priors over how the visual world moves, interacts, and responds to contact, making them promising substrates for robotic world modeling. The central challenge is how to communicate action to such models in a form aligned with the visual space in which they learned these interaction priors, yet still grounded in physical manipulation. We introduce Masked Visual Actions, a pixel-space control interface that expresses action as a partially revealed trajectory of an arbitrar

Reading and Steering Representations of Materials-Science Mechanisms in an Open-Weight Language Model huggingface.co

Large language models can answer scientific questions, yet a correct output does not reveal whether the model represents or uses the governing physics. Here we show that materials science mechanism information in the open-weight google/gemma-4-E4B-it model has three experimentally separable forms: concepts are readable in individual hidden states, constitutive orientation is carried by controlled transformations between states, and selected internal representations causally control engineering a

Where Should Optimizer State Live? Tiered State Allocation for Memory-Efficient Mixture-of-Experts Training huggingface.co

SkewAdam treats a mixture-of-experts model’s three parameter populations differently rather than storing identical AdamW moments for each. On a 6.78B-parameter MoE, standard AdamW burns 50.6 GB of first and second moments to update just 12.6 GB of bfloat16 weights, making optimizer state the top memory line item.

Subliminal Clocks: Latent Time Modelling in Diffusion Language Models huggingface.co

Diffusion language models are not conditioned on a timestep, yet probes recover a latent clock inside their residual streams. The signal encodes denoising progress and is read by downstream layers, showing DLMs learn to self-locate along the diffusion trajectory without being told where they are.

ISO: An RLVR-Native Optimization Stack huggingface.co

ISO reframes reinforcement learning with verifiable rewards around the singular structure of model weights, finding that RLVR reuses the base model’s spectra while shifting associated subspaces. The optimizer stack targets those subspaces directly, treating RLVR as a spectral edit rather than generic gradient descent.

Stale but Stable: Staleness-Adaptive Trust Regions for Stabilizing Asynchronous Reinforcement Learning huggingface.co

Asynchronous RL boosts throughput by decoupling rollouts from optimization, but policy lag, engine delays and MoE routing pile up staleness that PPO clipping only weakly constrains. The SAT method scales trust regions by measured training-inference divergence, treating clipping as a sampled surrogate and enforcing a full-policy bound.

AgentDebugX: An Open-Source Toolkit for Failure Observability, Attribution, and Recovery in LLM Agents huggingface.co

AgentDebugX targets a core pain of agent debugging: the failing step is rarely the causal one. The open-source toolkit wraps execution traces in a Detect, Attribute, Recover, Rerun loop, with a DeepDebug module running multi-turn root-cause diagnosis over full trajectories before patching and replaying.

AutoIndex: Learning Representation Programs for Retrieval huggingface.co

AutoIndex leaves retrievers and rerankers alone and instead searches over programs that slice, enrich, normalize and reweight documents prior to indexing. Agents diagnose failures on a validation set each iteration and propose new representation programs, treating retrieval quality as a compiler-style optimization target.

H^2SD: Hybrid Hindsight Self-Distillation huggingface.co

RLVR typically hands an entire reasoning trajectory one scalar reward, starving token-level credit assignment. H²SD borrows on-policy distillation’s dense token supervision but drops the external teacher, distilling instead from the model’s own successful hindsight rollouts to fill in per-token targets for math and code.

ABot-World-0: Infinite Interactive World Rollout on a Single Desktop GPU huggingface.co

We present ABot-World-0, an action-conditioned video world model for real-time, long-horizon closed-loop interaction, supported by a multi-source data infrastructure spanning AAA games, simulation engines, and internet videos to learn controllable world dynamics. WorldExplorer performs agent-driven collection guided by training feedback, while a unified pipeline applies 14 deterministic quality checks, VLM-based assessment, and synchronized action and text annotation. We progressively distill a

DataFlow-Harness: A Grounded Code-Agent Platform for Constructing Editable LLM Data Pipelines huggingface.co

Large language models (LLMs) are increasingly used to automate data-processing workflows, yet coding agents typically produce scripts that are not automatically materialized as persistent, editable platform artifacts. We call this disconnect the NL2Pipeline gap. To bridge it, we introduce DataFlow-Harness, a platform that guides an LLM agent to construct platform-native directed acyclic graphs (DAGs) through typed, incremental mutations rather than free-form scripts. The platform combines DataFl

HPD-Parsing: Hierarchical Parallel Document Parsing huggingface.co

Efficient teamwork typically combines global coordination with parallel execution, a principle not yet fully reflected in unified Vision-Language Model (VLM)-based document parsers. Existing unified parsers process an entire page jointly but generate its output through a single token-by-token autoregressive trajectory, creating a sequential bottleneck that grows with document length. Such full-page sequential generation overlooks a key property of document parsing: layout must be analyzed global

AlayaWorld: Interactive Long-Horizon World Modeling — Full Technical Report huggingface.co

Unlike conventional video game development, which relies on labor-intensive pipelines for asset production, animation, physics, and programming, video world models generate interactive environments from user inputs instantly. It enable us to create customized, explorable, and continuously evolving virtual world from text, an image, or video. Realizing this vision requires four tightly coupled capabilities: interaction, persistent spatiotemporal consistency, stable long-horizon generation, and ef

Generative World Renderer at the Speed of Play huggingface.co

Generative world renderer AlayaRenderer receives structured world states exported from physics engines and synthesizes RGB frames. Unlike models that generate frames from text/control-hints prompts, AlayaRenderer preserves scene structure without altering the underlying world dynamics. This demonstrates an alternative path toward interactive world modeling and user-controllable play. However, the original AlayaRenderer is too computationally expensive for real-time deployment. This technical rep

Two-Level Meta-Rubrics for Evaluating Open-Ended Generation: GAMUT, a Benchmark for Factual Completeness huggingface.co

Evaluating the factuality of long-form generations has focused predominantly on precision, measuring whether the claims a model makes are correct. The dominant decompose-search-verify pipeline catches incorrect claims well but says little about whether a response contains all the information it should. Measuring factual completeness, the missing half of factuality, is harder: it requires enumerating the full set of facts a complete answer should contain, and these facts rarely form a flat list.

Text Template Tokens Are Implicit Semantic Registers in Diffusion Transformers huggingface.co

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

Transcription Policy as a Latent Variable: Activating Controllable Verbatim ASR with Word-Level Timing huggingface.co

Modern ASR models trained on heterogeneously annotated data treat transcription style (verbatim vs. intended) as an uncontrolled latent variable, causing measurable decoding instability, evaluation confounding (up to 60% of reported WER attributable to style mismatch), and unreliable word-level timing. We show that models already encode both styles; the challenge is controlled activation. Using coverage-aware decoder task tokens trained on parallel verbatim/intended transcript pairs, we raise Ge

Delineate Anything v2: A Global Foundation Model for Field Delineation huggingface.co

Accurate agricultural field boundary delineation at large scale is a foundational task for food security, supply chain transparency, and carbon accounting. While vision foundation models like SAM show remarkable zero-shot capabilities, they frequently fail in geospatial domains due to topological complexity, cropland texturing patterns, and a lack of physical scale awareness. In this work, we introduce Delineate Anything v2, a globally scalable foundation model designed specifically for wide-are

Mage-Flow: An Efficient Native-Resolution Foundation Model for Image Generation and Editing huggingface.co

Large-scale visual generators are increasingly capable but costly to train, fine-tune, and deploy. We introduce Mage-Flow, a compact 4B-scale generative stack for efficient text-to-image generation and instruction-based image editing. The stack is built from two co-designed components: Mage-VAE, a lightweight high-fidelity latent tokenizer, and a Native-Resolution Multimodal Diffusion Transformer trained with rectified flow matching. Mage-VAE uses one-step diffusion-style encoding and decoding w

ConsiSpace: Learning Geometric Consistency Matters for Video Spatial Reasoning huggingface.co

Video spatial reasoning is essential for navigation-oriented perception and long-video question answering, where models must infer spatial relations across long horizons under changing viewpoints. However, existing multimodal large language models (MLLMs) remain largely semantic-centric, and often fail to reliably aggregate consistent spatial evidence from redundant video observations, leading to inefficient or unstable reasoning. To address these issues, we propose ConsiSpace, a geometry-consis

Appearance Pointers — Multimodal Region Control of Diffusion Transformers huggingface.co

Controllable image generation remains challenging for creative professionals, who often require precise regional control over materials, object identities, and spatial arrangements that cannot be reliably achieved through text prompting alone. Diffusion Transformers (DiTs) can natively ingest heterogeneous tokens stemming from texts and images, but they lack mechanisms for determining where and how these tokens should influence the output. We introduce appearance pointers, compact tokens that gu

SciForma: Structure-Faithful Generation of Scientific Diagrams huggingface.co

Structural fidelity is essential to scientific methodology diagrams. To communicate research logic, these diagrams must faithfully render components, directional relations, and textual annotations. Since a single error, such as a reversed arrow or an unreadable equation, can invalidate the entire figure, structural fidelity is inherently conjunctive: correctness on one axis cannot compensate for failure on another. Current open-source models fail to satisfy this criterion. Supervised fine-tuning

EduPanel: A Three-Agent LLM Judge for Teaching Videos — Reliability, Complementarity, and Human Trust Calibration huggingface.co

Teaching videos are becoming a major medium for education, creating a growing need for scalable evaluation of their pedagogical quality. Existing automatic judges do not fully address this setting because teaching quality depends on multimodal evidence and should be evaluated with respect to the intended learner rather than as a universal property. We present EduPanel, a rubric-grounded, learner-conditioned LLM judge that decomposes evaluation across specialized agents to produce interpretable a

Computational Humor with Multimodal LLMs: Methods, Datasets, Evaluation, and Challenges huggingface.co

Multimodal humor in memes, cartoons, and comics remains difficult for AI systems because intended meaning depends on non-literal mechanisms, shared cultural knowledge, and communicative intent rather than literal scene description. This survey focuses on visual humor understanding in single-image and multi-panel artifacts, while treating humor generation as an emerging downstream frontier. We position the literature against prior humor, sarcasm, and general MLLM surveys and organize it using a c

Trajectory-aware Cross-view Geo-localization with Sequential Observations huggingface.co

Cross-view geo-localization matches ground-level observations against geo-tagged satellite imagery. Recent methods show that sequential queries such as video clips yield richer spatiotemporal cues than single images, yet they overlook a complementary sequential modality: route descriptions — which capture the same trajectory at a higher level of abstraction and are often the only input available (e.g., a user directing an autonomous vehicle to a pickup point). To bridge this gap, we introduce S

References

CyberScoop cyberscoop.com

NIST mathematician Daniel Apon confirmed the results independently within hours of the public post, noting that the mathematics ‘checked out’; the HAWK authors were notified in June 2026 and helped validate the proof-of-concept before release.

RockCyberMusings weekly digest rockcybermusings.com

An anonymous team released findings showing they had achieved a similar result against HAWK using GPT-5.6, albeit through a different mathematical path — parallel discovery suggesting the era of AI-accelerated cryptanalysis is not vendor-specific.

Schneier on Security blog schneier.com

Schneier cautioned that the ‘scary’ framing of Mythos is partly a PR play… smaller, open-source models can reproduce similar results with sophisticated prompting, and researchers from AISLE recovered 100% of the analysis for certain flagship exploits.

PostQuantum.com analysis postquantum.com

Markku-Juhani Saarinen argued that AI-driven claims should be accompanied by machine-checkable proofs to prevent the community from being overwhelmed… human researchers now spend hundreds of hours verifying what an AI produces in days.

AI Weekly aggregation aiweekly.co

To restore original security targets HAWK would likely need to double its key sizes, a move that would eliminate the compactness advantage that made it a competitive candidate against Falcon and ML-DSA.

Matthew Green / Decipher release notes aiweekly.co

Green highlighted a critical agentic risk: left to their own devices, AI agents often attempt to download and compile unverified, third-party repositories to solve cryptanalysis tasks — arguing AI functions best as a force multiplier for human experts rather than a fully autonomous replacement.

Ctrl-World project page ctrl-world.github.io

Ctrl-World introduces frame-level action conditioning with pose-conditioned memory retrieval, maintaining long-horizon consistency for over 20 seconds and improving real-world policy success from 38.7% to 83.4% via synthetic rollouts.

1X Technologies — World Model Self-Learning 1x.tech

1XWM uses a 14B video backbone that ‘hallucinates’ futures which an Inverse Dynamics Model translates into actuator commands for the NEO humanoid, pre-trained on internet video then mid-trained on 900 hours of egocentric human data before 70 hours of NEO logs.

Seer: Predictive Inverse Dynamics Models (arXiv 2412.15109) arxiv.org

Seer generates ‘visual foresight’ tokens to reconstruct future RGB images that guide an inverse dynamics module to speculate intermediary actions, yielding a 43% improvement in real-world tasks and new SOTA on CALVIN ABC-D.

Video Language Planning (VLP) project page video-language-planning.github.io

VLP uses a tree search where VLMs serve as both policy and value functions and text-to-video models act as the dynamics engine — a precedent for MVA’s Best-of-N with Gemini 1.5 Pro as critic.

ICRA 2025 Workshop: Beyond Pick-and-Place sites.google.com

Robots remain plateaued at pick-and-place behaviors, struggling with the contact-rich skills necessary for industrial assembly or cooking; visual common sense does not account for force-torque requirements of real-world manipulation.

EmergentMind — DROID dataset overview emergentmind.com

DROID comprises ~76,000 teleoperated trajectories across 564 scenes but is exclusively single-embodiment (Franka Panda 7-DoF); zero-shot generalization to very different manipulators still often requires additional data or transfer techniques.

Anthropic — ‘A Global Workspace in Claude’ (transformer-circuits.pub, July 2026) transformer-circuits.pub

The Jacobian lens computes the linearized effect of an internal activation on the model’s future likelihood of producing specific tokens, surfacing ‘verbalizable representations’ the model is poised to express but has not yet written.

daily.dev summary of Anthropic Jacobian lens work daily.dev

Fitting the lens can take hours, and full-vocabulary readouts during deployment can increase inference costs by up to 400%… on smaller models the J-lens may actually underperform the simpler logit lens unless a specific one-parameter correction is applied.

Erik Hoel, The Intrinsic Perspective theintrinsicperspective.com

The J-lens might merely be a highly sophisticated tracker of output probability rather than a window into ‘thought’… the field remains in a pre-paradigmatic state where mathematical success is often mistaken for philosophical insight.

nostalgebraist, Alignment Forum — ‘Interpreting GPT: the Logit Lens’ alignmentforum.org

The logit lens assumes the model’s internal ‘basis’ aligns with the output space, which is rarely true in early layers — yielding garbled or misleading intermediate readouts.

lamm-mit/Substrates GitHub repository github.com

Repository ships the gemma4-materials-mechanism-prompts corpus spanning 21 dataset configurations, the 60-law counterfactual benchmark, and the 72-prompt answer-scaffold graph for blinded mechanism identification.

AI Agents Plus — ‘Mechanistic Interpretability, MIT 2026 Breakthrough’ ai-agentsplus.com

MIT Technology Review recognized mechanistic interpretability as a 2026 Breakthrough Technology, but critics highlight the ‘interpretability illusion’ where clean-looking circuits fail to generalize across datasets.