AI Research news and analysis on Web Pulse: 8272 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
Researchers at AE Studio and Anthropic developed GRAM, a method that confines dual-use knowledge in AI models to removable modules, allowing surgical removal of dangerous capabilities without retraining. The technique ad…
Anthropic, in collaboration with AE Studio, introduced Gradient-Routed Auxiliary Modules (GRAM) on July 8, 2026, a pretraining method designed to restrict access to dual-use knowledge in AI models by routing sensitive ca…
A new paper presented at ICML introduces Latent Collaboration in Multi-Agent Systems (LatentMAS), enabling AI agents to share latent states directly instead of text, boosting performance at the cost of interpretability. …
Daniel Hulme, CEO of Conscium and chief AI officer at WPP, asked Elon Musk on a yacht in Cannes to invest in solving machine consciousness. Hulme, a leading UK AI expert, discussed his career and vision for building curi…
OpenAI retired Advanced Voice Mode on July 8 and replaced it with GPT-Live, a full-duplex voice model that processes input while generating output, enabling natural interruptions and backchannels. The system delegates co…
Grok 4.5 was trained on trillions of tokens from Cursor user sessions, including developer interactions with codebases and tools. The model shows strong results on workflow benchmarks and token efficiency, likely due to …
A new research paper argues that generative AI struggles with static malware analysis because a single malicious file can contain more data than an entire AI training dataset, making it difficult for models to learn effe…
Researchers propose EvoSOP, a framework enabling LLM agents to synthesize atomic actions into reusable Standard Operating Procedures (SOPs) for self-evolution. Experiments show EvoSOP significantly improves task success …
Researchers developed cost-effective agent harnesses for the ARC-AGI-1 abstract reasoning benchmark, achieving 67.25% pass@2 at $0.62 per task using an open-weight model without fine-tuning. The Reflective Orchestrator a…
Researchers propose MILES, a framework that enables large language models to accumulate and reuse step-wise reasoning experience across sequential problems, improving accuracy and efficiency without requiring large-scale…
Researchers introduced SpaR3D-MoE, a framework that enhances multimodal large language models with adaptive 3D spatial reasoning from sparse RGB inputs. It uses a spatiotemporal manifold sampling mechanism and a geometry…
Researchers demonstrate that state-of-the-art video diffusion models possess a highly structured latent space capable of both generation and understanding. They introduce Gen4U, a framework that repurposes frozen generat…
A new study from Writer finds that orchestration design, specifically the Writer Agent Harness, reduces enterprise agentic AI costs by 41% and tokens per task by 38% while maintaining quality, outperforming model swaps a…
Researchers developed a multi-analyst LLM pipeline that converted 68 public physiological corpora into 436 unique candidate rule shapes for contactless monitoring detectors, with 94 build-now components identified after …
Researchers introduced the Large Behavioral Model (LBM), a promptable digital twin that learns customer decision-making from retail transactions. The model outperforms general-purpose language models on in-domain retail …
Researchers propose D2PO, a framework that optimizes diffusion sampling policies by reformulating sampler optimization as a preference-based alignment problem using Direct Preference Optimization (DPO). D2PO introduces d…
Researchers propose Task-Semantic Field Factorization (TSF), a large language model-guided framework that integrates variable semantics from process documents into time-series forecasting for industrial processes. TSF re…
Researchers propose ROAM, a framework that uses LLM reasoning to adapt frozen specialist models to new scenarios without retraining, reducing MAE by over 20% in industrial processes with minimal overhead.
Researchers introduced Physics-Audited Agentic Scientific Machine Learning (PA-SciML), a verification-first workflow that ensures surrogate models discovered by LLM agents satisfy physics constraints such as boundary con…
Researchers introduced STST-JEPA, a self-supervised transformer for EEG analysis, pretrained on 47,703 sessions from ages 5-81. The model achieved a mean absolute error of 3.06 years in age regression and ranked first on…