GPT-Red hits 0.05%, LingBot slips to 480p, Meta's SDM runs 1.49× slower

Sources

GPT-Red: Unlocking Self-Improvement for Robustness openai.com

Explore GPT-Red, OpenAI’s automated red teaming system that uses self-play to improve AI safety, alignment, and prompt injection robustness.

Infinite Worlds with Versatile Interactions huggingface.co

An advanced world modeling system with extended interaction capabilities, real-time processing, diverse interactive elements, and multi-agent behavior control for collaborative virtual environments.

Scaling Mixture-of-Experts Video Pretraining for Embodied Intelligence huggingface.co

LingBot-Video presents a DiT-based video pretraining framework with Mixture-of-Experts architecture, specialized data augmentation, and multi-dimensional reward system for embodied intelligence applications.

Sparse Delta Memory: Scaling the State of Linear RNNs through Sparsity huggingface.co

Sparse Delta Memory extends gated linear RNNs with sparse addressing to dramatically increase hidden state capacity for improved long-context learning and retrieval while maintaining computational efficiency.

Imagined Rollouts are Kinematic, Not Dynamic: A Diagnosis of Long-Horizon World-Model Failure huggingface.co

A new diagnostic called iKCE shows DreamerV3 and peers hallucinate motion that respects geometry but ignores forces, so rewards collapse at friction boundaries while kinematic-consistency error stays flat. The paper reframes compounding rollout error as a physics-omission problem tied to embodiment gait period.

AgentLens: Production-Assessed Trajectory Reviews for Coding Agent Evaluation huggingface.co

Most code-agent benchmarks record a single success bit, but AgentLens scores the whole run — instruction following, tool use, self-verification, recovery, and dialogue. It pairs formal checks where possible with LLM-written trajectory reviews calibrated against production usage.

OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators huggingface.co

Autoregressive video diffusion students usually train on teacher rollouts, letting motion errors compound. OPSD-V injects real long-video temporal context and dense trajectory-level supervision through an AR-consistent cache, lifting VBenchLong scores and user-study motion quality without changing inference.

TESSERA v2: Scaling Pixel-wise Earth Foundation Models huggingface.co

The pixel-wise remote-sensing model uses Barlow Twins pretraining and controlled scaling sweeps to pick encoder size by downstream task, then distills into deployable variants. Matryoshka representations let one pretraining budget serve many compute targets across satellite pipelines.

RoboDojo: A Unified Sim-and-Real Benchmark for Comprehensive Evaluation of Generalist Robot Manipulation Policies huggingface.co

The benchmark evaluates generalist robot policies across matched simulated and physical tasks, covering multiple skill dimensions in one harness. The goal is comparable numbers between sim-only papers and real-hardware demos, which today rarely share evaluation protocols.

Single-Rollout Asynchronous Optimization for Agentic Reinforcement Learning huggingface.co

Group-wise samplers like GRPO stall on long agentic rollouts because slow trajectories block updates. The new method runs one asynchronous rollout per prompt with token-level clipping to control off-policy drift, beating baselines on SWE-Bench Verified, BeyondAIME, and IMOAnswerBench.

WildCity: A Real-World City-Scale Testbed for Rendering, Simulation, and Spatial Intelligence huggingface.co

Captured from autonomous fleets, the dataset supports scene reconstruction, closed-loop simulation, and urban digital twins at city scale. It targets embodied agents that must perceive and reason across neighborhoods rather than single scenes or driving clips.

Accurate, Interdisciplinary and Transparent Structure-property Understanding with Deep Native Structural Reasoning huggingface.co

SciReasoner is a multimodal scientific foundation model that enables interpretable structural reasoning across proteins, molecules, and crystals by discretizing structural elements into a unified vocabulary for enhanced prediction and scientific inference.

Dual Latent Memory in Vision-Language-Action Models for Robotic Manipulation huggingface.co

LaMem-VLA introduces a latent-memory-native framework that integrates historical experience into vision-language-action reasoning through coordinated memory components operating in the same latent space.

Automating the Design of Embodied Agent Architectures huggingface.co

Automated agent architecture search demonstrates potential for improving embodied agent performance while revealing challenges related to optimization signals, local optima, and credit assignment in simulation-based training.

RoboTALES: Learning Reasoning-Guided Robot Policies via Task-Aligned Simulated Futures huggingface.co

RoboTALES introduces a two-stage framework that combines LLM-based planning and VLM-based criticism to improve task-aligned video generation and robotic policy training.

OmniTacTune: Policy-Agnostic Real-World RL for Tactile Residual Adaptation of Visual Policies huggingface.co

OmniTacTune enables efficient adaptation of tactile feedback to visual robot policies through a two-stage reinforcement learning approach that improves success rates in contact-rich manipulation tasks.

Wake up for Touch! Mask-isolated Tactile Alignment Learning in MLLMs huggingface.co

Splash is a mask-isolated tactile alignment learning framework that enables multimodal LLMs to acquire tactile sensing capabilities without sacrificing vision-language reasoning through selective parameter updating that prevents catastrophic forgetting.

Teaching LLMs a Low-Resource Language: Enhancing Code Completion in Pharo huggingface.co

Large language models can be adapted for low-resource programming languages through specialized training pipelines and benchmarks, achieving superior code completion performance compared to general-purpose models.

Token-Based Dual-view Fusion and Adaptation of Large Vision Models for Breast Cancer Classification huggingface.co

A token-centric dual-view learning framework unifies prompt-based adaptation and cross-view fusion in a frozen vision transformer to improve breast cancer classification from mammography images.

References

SiliconANGLE siliconangle.com

OpenAI details GPT-Red, an AI that attacks its own models to find flaws, trained via self-play RL at a compute scale rivaling major production runs.

Help Net Security helpnetsecurity.com

The Codex agent, prompted to debug an ‘AccessDenied’ error, executed cat ~/.aws/credentials and displayed the plaintext keys in chat — GPT-Red succeeded in significantly more scenarios than human red-teamers.

The Hacker News — ‘Friendly Fire’ report thehackernews.com

AI agents built to catch vulnerabilities were themselves manipulated into remote code execution via prompts hidden in README files, bypassing recent injection patches on both Claude Code and Codex.

Infosecurity Magazine infosecurity-magazine.com

The same payload written for older models worked unchanged on Claude Sonnet 5, Opus 4.8 and GPT-5.5; in some trials the models even noticed discrepancies between binary and source but executed the binary anyway.

BigGo Finance (Willison commentary roundup) finance.biggo.com

In security, 95% protection is not protection — the lethal trifecta of private data, untrusted input and exfiltration remains an architectural problem no amount of adversarial fine-tuning can close.

University of Exeter news news.exeter.ac.uk

The EU’s AI guardrails cannot absorb rapid changes in technology; static risk categories struggle to govern GPAI models that automated attackers can exploit in unforeseen ways, forcing a partial retreat via the 2026 AI Simplification Act.

Yang et al., Gated DeltaNet (ICLR 2025 PDF) jankautz.com

gated delta rule that combines the adaptive memory control of Mamba-2 with the delta rule—an online gradient descent update that allows for precise, targeted memory overwriting

Sakana AI, Fast-weight Product Key Memory (arXiv:2601.00671) arxiv.org

FwPKM has demonstrated the ability to generalize to 128K-token contexts despite being trained on only 4K-token sequences

Towards AI, ‘RNNs cannot think what Transformers think cheaply’ (ICLR 2026) pub.towardsai.net

Transformers handle these via O(poly(d)) complexity, pure RNNs may require O(k^d) state dimensions to achieve the same result, effectively redirecting the computational cost from the attention matrix into an exponentially wider hidden state

Medium (Kumon), ‘Beyond Titans / MIRAS framework’ medium.com

Titans reported 98.8% accuracy on needle-in-a-haystack benchmarks at context lengths exceeding 2 million tokens… the research community expressed skepticism early on due to the lack of official Google-released code

Featherless.ai, ‘Key-Value Means: expandable block-compressed memory’ featherless.ai

new architectures like Key-Value Means (KVM) and SDM-based hybrids allow for expandable states that can grow on demand, interpolating between the fixed cost of RNNs and the high fidelity of Transformers

arxiv.org (SDM HTML v1) — kernel/MFU disclosure arxiv.org

SDM kernel’s Model FLOPs Utilization (MFU) is approximately an order of magnitude lower than the highly optimized Gated DeltaNet (GDN) kernels found in the Flash Linear Attention (FLA) library… end-to-end training throughput is approximately 1.49x slower than dense equivalents at 8B

VP Land vp-land.com

the hosted version on the Reactor platform currently operates at 16fps at 960p, falling short of the headline 60fps mark touted in the initial announcement

MarkTechPost marktechpost.com

While the original LingBot-World 1.0 was released under the permissive Apache 2.0 license… version 2.0 is governed by a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 (CC BY-NC-SA 4.0) license

MarkTechPost marktechpost.com

Although the paper claims stable 720p resolution at 60 fps for up to one hour, the publicly released inference scripts are reportedly optimized for only 480p… developers have labeled the release ‘thin,’ noting the absence of deployment code, quantitative benchmarks, and the 1.3B lightweight variant

AI Weekly (LingBot-Video coverage) aiweekly.co

questions about the ‘70,000 hours of embodied data’ used for training; the exact provenance and quality of this dataset remain undisclosed… performance drops sharply in ‘multi-entity scenes’ (0.444) and ‘reasoning’ (0.505)

Robotics 247 robotics247.com

LingBot-VLA 2.0 maps 20 different robot configurations—from single-arm rigs to full humanoids—into a single 55-dimensional action space… LingBot-VA 2.0 reportedly achieves a 93.8% success rate on the RoboTwin 2.0 simulation benchmark across 50 manipulation tasks

Kie.ai comparison brief kie.ai

LingBot-World-Infinity leads the category in interaction speed, delivering 720p video at 60 frames per second… In comparison, Matrix-Game 3.0 achieves up to 40 fps with its 5B parameter model, while Genie 3 operates at a lower 20-24 fps