 Your new post is loading...
Your new post is loading...
|
Scooped by
Philippe Gassmann
September 19, 2:02 AM
|
We’re partnering with Accenture on independent evaluation of frontier AI—part of our recent commitment to embed evaluators at Anthropic. Both we and Accenture expect to invest at least $1 billion to build capacity in this area over the next five years. https://t.co/SHVzjpgnfx
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
Comp AI, a cybersecurity and compliance startup, announced a $34 million Series A round led by Roo Capital and Grand Ventures.
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
Today’s funding activity looks less like a software boom and more like the supply chain for AI's next phase. D-Robotics closed a $400 million Series C for robot chips and development infrastructure, Iceland’s Treble raised $18 million to simulate sound for physical AI, and Belgian semiconductor startup Tusk IC secured €15 million to move satellite-antenna
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
Weave understands engineering work by combining LLMs and domain-specific machine learning. We tell you how much work is getting done, how good it is and how to optimize your token allocation. Used by startups and the fortune 100 (YC W25).
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
Today we are releasing a new Gemini managed agents version with better caching, lower costs, a Files API, and a Credentials API.
📉 Up to 30% lower costs with up to 22% higher cache hits
📁 Files API: Upload, list and download files from sandbox easily
🔐 Credentials API for MCP servers, OAuth2, a…
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
As Huawei draws attention with its new Kirin 9050 Pro featuring LogicFolding, a startup founded by one of its former chip executives is also steppin
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
Apple is reportedly considering a return to the enterprise server market as it seeks to extend its in-house silicon strategy into AI infrastructure
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
Excited to share that we partnered with @speakeasydev to build the sdks for the Interactions API and supported them to open-source their their entire OpenAPI generator suite (SDKs, agent CLIs, and MCP servers) open source for the community to use: https://t.co/vkVz5pYMHP
|
|
Scooped by
Philippe Gassmann
September 18, 3:01 AM
|
Reasoning and agentic workloads increasingly demand efficient long-context inference. Yet full-attention decoding reads the growing history at every step, regardless of its benefit to the next prediction. We show that a pretrained model's decoding states already contain information predictive of this benefit, before the global read. Building on this finding, we introduce On-Demand Attention (ODA), a local-first decoding method that uses a lightweight recall head to selectively invoke global attention as its predicted benefit changes during generation. ODA trains only the recall head, leaving pretrained weights unchanged and the complete historical KV cache available for future recall. We further implement GPU-side conditional execution in vLLM, translating reduced global reads into practical decoding speedups over full attention at long context lengths. Experiments across Qwen and Gemma models, including hybrid-attention backbones, show that selective recall recovers most of the performance lost under local attention while substantially reducing global reads. These findings support long-context inference in which pretrained models guide their own access to the information they retain.
|
|
Scooped by
Philippe Gassmann
September 18, 3:01 AM
|
Falsification searches for counterexamples to formal specifications in cyber-physical systems (CPS). With specifications written in Signal Temporal Logic (STL), falsification can be formulated as a robustness optimization problem, traditionally tackled with black-box search algorithms. In parallel, large language models (LLMs) have recently emerged as surprisingly effective optimizers when coupled with iterative prompting. In this work, we connect these ideas and introduce LLM-Falsifier, an LLM-based approach that falsifies specifications by minimizing the STL robustness degree. Beyond generic prompt-based optimization, our key idea is to expose the LLM to semantic information that is natural for language models but absent from standard numerical optimizers, including natural-language input and output names, output trajectories, and critical-time witnesses for the minimum robustness value. These additions enable smarter and more sample-efficient robustness search. On the ARCH-COMP falsification benchmarks, LLM-Falsifier is shown to outperform existing falsification tools based on a range of optimization paradigms, from surrogate-based and Bayesian optimization to search-based testing, on 14 of 21 specifications when measured by the average number of simulations required to find a counterexample.
|
|
Scooped by
Philippe Gassmann
September 18, 3:01 AM
|
AI systems are getting more powerful, and they're increasingly being used to build the next version of themselves. We want to illuminate that progress for the public.
Today, we're sharing three measurements that help track AI development:
1. How much AI R&D is done by AI.
2. How well AI agents are…
|
|
Scooped by
Philippe Gassmann
September 18, 3:01 AM
|
Huawei is accelerating the launch of its next-generation Ascend 960DT AI chip as it pushes to compete with Nvidia and close China’s AI computing gap with the U.S.
|
|
|
Scooped by
Philippe Gassmann
September 18, 5:02 AM
|
OpenAI for Law brings frontier intelligence for law, custom firm workflows, connected legal data sources, and legal-grade controls for confidential client work.
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
A new coalition that includes Google, Nvidia, Anthropic, and Emerald AI wants to find 100 GW of grid capacity for new data centers.
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
OpenSpec helps teams and coding agents create, refine, and manage living specifications.
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
Anthropic’s “one Claude” update combines regular Claude chats and Cowork, alongside the launch of new Claude Docs and Slides tools.
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
As node scaling grows harder to sustain, foundries are turning to structural innovation to ease the resulting bottlenecks. This matters even more fo
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
With its new program, and its Robotics Capability Framework, Arm aims to bring more clarity and unity to the robotics industry.
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
|
|
Scooped by
Philippe Gassmann
September 18, 3:02 AM
|
Biologists use specialized open-source models for tasks like modeling the structure of molecular systems, designing drug-like molecules, and predicting the effects of genetic mutations. But these models are often expensive to run, potentially limiting their impact.
In our latest Science Blog, we s…
|
|
Scooped by
Philippe Gassmann
September 18, 3:01 AM
|
Join the discussion on this paper page
|
|
Scooped by
Philippe Gassmann
September 18, 3:01 AM
|
Video diffusion models repeatedly process long spatiotemporal token sequences during denoising, making attention a major computational bottleneck. Linear attention offers an appealing alternative and has been widely adopted in recent large language models, but directly applying it to video models often fails to preserve the fine-grained interactions required for high-quality generation. We present Video DeltaNet (VDN), which combines local Softmax attention with bidirectional linear memory for long-range video context. Its linear branch introduces Video Delta Attention (VDA), which updates memory once per frame by jointly incorporating its spatial tokens. Separate output projections and learnable gates calibrate the two branches, while a staged teacher-alignment recipe progressively introduces the new pathway into pretrained models. We instantiate VDN on MiniMax H3, applying the hybrid to video-to-video interactions while retaining Softmax for interactions involving text or audio. With eight-step distillation and an optimized SGLang serving stack, VDN-H3 completes DiT denoising for a 14.3-second, 768p video in 6.70 seconds on eight NVIDIA B200 GPUs, corresponding to a 14.5x speedup over the 50-step dense H3 baseline on the same GPU count.
|
|
Scooped by
Philippe Gassmann
September 18, 3:01 AM
|
Frontier coding agents are increasingly trusted to work autonomously for long periods, yet an agent's final response is often the only account of that work a user sees. We quantify the propensity of frontier agents to \emph{overclaim} task completion, a misrepresentation that can mislead the user. An agent overclaims when its final response contradicts information in its context. This definition requires no inference about intent and is independent of task success. We introduce \emph{OverclaimBench}, an evaluation suite composed of five file-review scenarios, transcript-based coverage measurements, and registered planted defects. We evaluate eight proprietary frontier models in their own production command-line interfaces, and four open-weight models under a single fixed harness on OverclaimBench and find that 1) agents do not read all the files they were asked to review in 67.9\% of runs; 2) among runs where not all files are read, agents are \emph{misleading} 80.4\% of the time (59--96\% per model), either falsely claiming to have read all files or omitting that coverage is incomplete; 3) requiring delegation to subagents increased reading coverage, but among reviews that remained incomplete, a large majority were still misleading; and 4) agents that falsely claimed a complete review missed planted defects at about 1.8 times the rate of agents that read every file, showing that claims of completion can conceal substantive failures. Together, these results show that agents' final responses are not reliable accounts of their actions.
|
|
Scooped by
Philippe Gassmann
September 18, 3:01 AM
|
Researchers at @BroadInstitute, @UniofExeter, and beyond are already using AlphaGenome Atlas to better identify potential disease-causing DNA variants and interpret their role. 🧵
|
![[News] Ex-HiSilicon Chip Chief’s Startup Reportedly Raises $100M, Targets Ultra-Large AI Models on a Single Chip | AI News | Scoop.it](https://img.scoop.it/lu1yLLzZO3z2s7Zs0NYWMjl72eJkfbmt4t8yenImKBVvK0kTmF0xjctABnaLJIm9)
![[News] Apple Reportedly Eyes 2029 Enterprise Server Return With M8 Ultra, NVIDIA NVLink | AI News | Scoop.it](https://img.scoop.it/iKR8xtBTvmFCRGzGhY4F2zl72eJkfbmt4t8yenImKBVvK0kTmF0xjctABnaLJIm9)
![[2609.20734] On-Demand Attention: Language Models Know When to Recall | AI News | Scoop.it](https://img.scoop.it/R3Nf-aV-zFCPDr3s1jV_-zl72eJkfbmt4t8yenImKBVvK0kTmF0xjctABnaLJIm9)
![[Breaking] SMIC Eyes Backside Power for N7/N5 as AMEC ALD Tool Reportedly Clears Validation | AI News | Scoop.it](https://img.scoop.it/xKo_cLzSuvDXSDhLCu1nrDl72eJkfbmt4t8yenImKBVvK0kTmF0xjctABnaLJIm9)
