Industry Adoption C

An arXiv paper presents a systematic analysis of how LLMs can learn, via reinforcement learning, to adapt their query formulation strategies to different retrievers in retrieval-augmented generation. Summarized neutrally from the abstract.

A developer shares lessons from building an LLM-driven tool that answers user questions via a customer API. Capturing and writing down domain knowledge is much of the work, easier than earlier AI generations since it need not be rigidly structured, yet exactly where prior efforts foundered.

Gen-VCoT replaces text-only chain-of-thought with generated RGB intermediates, staging visual grounding (SAM), depth (Marigold), and semantic reasoning (Qwen2-VL) under an adaptive router. It improves spatial (+25%) and depth (+50%) questions but can hurt simple factual ones; text CoT still wins on CLEVR, suggesting task-dependent representations.

As LLM post-training RL uses multi-dimensional rewards, conflicting signals across reward groups can cancel out and hinder training. GD²PO decouples rewards into groups and, inspired by DAPO, dynamically filters near-zero-advantage rollouts, reducing conflicts and improving RL training efficiency.

This arXiv paper proposes agentic automata learning to assess how well tool-calling LLM agents can uncover hidden environments through interaction. An agent must infer a hidden deterministic finite automaton (DFA) via membership and equivalence queries, yielding a scalable testbed with controlled task complexity. Evaluating state-of-the-art LLMs, the authors find performance drops sharply as DFA size grows, with reasoning models markedly stronger.

While knowledge is managed via Wikipedia and software via GitHub, agent skills still lack infrastructure for large-scale production, governance, and evolution. SkillWiki is a living knowledge infrastructure turning heterogeneous knowledge into reusable skill assets linked to their originating evidence. It presents the full skill lifecycle, from knowledge ingestion to provenance-aware exploration, governance, and execution-driven evolution, with a live demo and source code available.

GPU kernel optimization assumes correctness and targets execution efficiency. The authors present daVinci-kernel, an RL framework coupling skill discovery and exploitation via a dynamically evolving skill library. Three agents share one LLM backbone: a Selection Agent retrieving techniques via BM25 and LLM reranking, a Policy Agent generating CUDA/Triton kernels, and a Summary Agent distilling rollouts into reusable skills. Skills are added only after execution verification confirms speedups.

IBM's new AI tool 'IBM Bob' reportedly cut Java application modernization from 30 days to 3 at early adopters. Its distinguishing feature is going beyond mere source-code generation.

Cohere announced it will move to a larger London office at 100 New Oxford Street, nearly tripling its UK footprint. The expansion backs the city's AI talent and R&D base and supports growing demand for secure, enterprise-grade sovereign AI across the UK and Europe.

OpenAI introduced its Partner Network, committing $150M to help global partners accelerate enterprise AI adoption, deployment, and transformation. The program aims to broaden OpenAI's reach into enterprise markets through a structured partner ecosystem.

Sakana AI has launched Sakana Marlin, its first commercial product: an autonomous research assistant for business. Given a research theme, it works autonomously for up to about eight hours—forming hypotheses, gathering and verifying information—then outputs structured summary slides and a report spanning dozens of pages. Built on the firm's long-horizon reasoning technology, it aims to act as a 'virtual CSO,' is self-serve, and available same day, with plans from free pay-per-use to Enterprise.

Simon Willison comments on the US government's national-security export-control directive suspending all foreign-national access to Fable 5 and Mythos 5, calling the move extraordinary and questioning its rationale and impact.

A hands-on test of using the new Fable 5 AI to produce a YouTube video. The author is impressed by the surprisingly high quality of the output but also flags a significant weakness in the workflow.

Anthropic announced a partnership with Tata Consultancy Services. TCS will deploy Claude to 50,000 employees across 56 countries, build Claude-powered products for finance, healthcare and the public sector, and join the Claude Partner Network.

AgentSpec studies scaffolded LLM agents that combine reasoning, memory, reflection, and action through controlled composition. It aims to isolate how each component contributes to overall performance.

Transformer-based ASR models like Whisper are accurate but hard to interpret, and existing XAI methods lack faithfulness and temporal precision. The paper proposes an entropy-guided explainability approach for such audio models.

Verifier-driven self-DPO, where a frozen verifier scores candidates to form preference pairs, is a common recipe for self-improving vision-language models. The paper shows that under this setup VLMs can regress on new tasks when the verifier misbehaves.

The study presents a statistical and machine learning framework characterizing a hydrogen-based multi-energy system. It targets operational threshold detection and deployable dispatch controller development.

LLM agents increasingly run as long-lived autonomous runtimes that schedule jobs, call tools, maintain memory, and push results to humans. This longitudinal study of one persistent system presents a taxonomy of its silent failures.

Applying GRPO to GUI grounding samples rollouts from a single screenshot, so groups often turn all-failure or all-success and yield weak signal. VISTA introduces view-consistent, self-verified training to stabilize GUI grounding.

Internet of Medical Things devices handle sensitive health data under tight resource constraints, making security and privacy critical, while federated learning adds complexity. The paper proposes an edge-native federated learning approach to secure IoMT in the post-quantum era.

NVIDIA's developer blog explains how to deploy MiniMax M3 on NVIDIA accelerated infrastructure for long-context reasoning and agentic workflows, addressing fragmented enterprise AI pipelines spanning text, vision, and other modalities.

Knowing how a model will perform in a new environment before deployment helps prevent harm. The paper evaluates predictions under two common sources of degradation: distribution shift and selection bias.