Official CMU (Carnegie Mellon AI) technical update and publication covering Carvalho Earns NSF CAREER Award To Study Motivation and Learning.
<div class="block-paragraph_advanced"><p><span style="vertical-align: baseline;">For the millions of fervent fans of the </span><a href="https://www.iplt20.com/" rel="noopener" target="_blank"><span style="text-decoration: underline; vertical-align: baseline;">Indian Premiere League</span></a><span style="vertical-align: baseline;"> (IPL), being able to count on a flawless live streaming cricket e
TorchServe is no longer maintained, leaving teams to own the entire GPU inference stack. The AWS Ray Serve Deep Learning Container is a supported, pre-tested container with the framework, GPU drivers, and serving layer already assembled. This post walks through deploying a vision-language model on Amazon EKS using the Ray Serve DLC on a single GPU node.
Amazon Quick custom permissions let you enforce least-privilege access by toggling features per user. This post walks through four patterns to automate custom permissions across the user lifecycle: a RegisterUser API parameter, account and role defaults, event-driven Amazon EventBridge and AWS Lambda automation, and a retroactive batch update script.
Official technical announcement and publication from Hugging Face covering IBM releases SOTA Granite Time Series PatchTST-FM-r2 model with commercial-friendly license.
The post Helping Legal Teams Own Their Intelligence: Our Series H Investment in Harvey appeared first on Lightspeed Venture Partners .
Chris Lehane argues that stronger AI capabilities require stronger safety evidence, shared standards, and durable policy action while the policy window remains open.
Modernizing complex legacy code with AI agents.
Meet GPT-6 Astra, OpenAIโs most capable model for business, with advanced reasoning, computer use, and stronger writing and design judgment.
Official Deepgram technical update and publication covering Deepgram Launches Flux Multilingual: The Worldโs First Multilingual Conversational Speech Recognition Model.
We construct explicit finite Gabor windows whose orbits are simultaneously full spark and phase retrievable. In every cyclic dimension, algebraic specialization preserves all Gabor minors while removing every zero of the self-ambiguity function, thereby resolving the explicit simultaneous- construction problem arising in finite nilpotent-group phase retrieval. We isolate the underlying regularizat
Research & Technology
Official Harvey AI technical update and publication covering We Raised $550M at a $15.5B Valuation to Help Legal Teams Own Their Intelligence Learn more We Raised $550M at a $15.5B Valuation to Help Legal Teams Own Their Intelligence Learn more.
Official technical announcement and publication from Together AI covering The Open Source AI Stack.
Press releases and partnership announcements.
Soniox Text-to-Speech now includes more than 200 built-in voices, tagged by gender, age, accent, use case, and style, and every voice speaks 60+ languages.
Official Mistral AI technical update and publication covering Modernizing complex legacy code with AI agents..
Sep 9, 2026
Official Harvey AI technical update and publication covering How to Use Legal AI to Produce Red Flag Reports For Due Diligence.
Official Harvey AI technical update and publication covering How In-House Legal Coordinates Compliance Across Regulations, Contracts, and Policies.
Official Harvey AI technical update and publication covering How to Use Legal AI to Produce Red Flag Reports For Due Diligence.
Posted on September 9, 2026
On June 3, 2026, Gradium's time to first audio on the Coval TTS leaderboard moved from 171.9 ms to 429.6 ms. No model had shipped and no serving infrastructure had changed. What changed was the measurement: Coval had started counting the silence at the start of the stream.
How Devin and a Cognition researcher built the worldโs highest-performance GPU lattice siever, to make factoring numbers 10x cheaper than the previous state of the art
Gradium Text-to-Speech is now available on LiveKit Inference, free until Oct. 9. Gradium TTS is designed for conversational use cases, with voice cloning, low latency (below 250ms TTFA), and robust pronunciation on difficult cases.
Official Harvey AI technical update and publication covering How In-House Legal Coordinates Compliance Across Regulations, Contracts, and Policies.
Official Harvey AI technical update and publication covering We Raised $550M at a $15.5B Valuation to Help Legal Teams Own Their Intelligence Learn more We Raised $550M at a $15.5B Valuation to Help Legal Teams Own Their Intelligence Learn more.
GPT-6 Astra from OpenAI is now generally available on Amazon Bedrock. It brings deeper reasoning and sharper judgment to your most demanding tasks, running on the Amazon Bedrock inference engine built for high performance, security, and scale.
288 points, 297 comments
Today weโre open-sourcing hyper-๐-bench, a new long horizon agent evaluation that measures how well models can not only act as an agent, but construct one.
Foundation vision-language models (VLMs) exhibit broad intelligence about the world, yet translating this intelligence into robot control remains challenging. We present Show-Harness, an Embodied Harness that enables VLMs to "play" robots through a compact semantic interface linking intent to action. Show-Harness exposes discrete semantic action units that VLMs can naturally reason over, while emb
We introduce MetroLLM-Bench, a 955-case benchmark for testing language models as the policy layer of a transit kiosk. It covers six real metro systems, ranging from 37 to 414 stations, and eleven categories that include routing, fare calculation, disruptions, accessibility, and adversarial input. In each case, the model must call structured tools and submit a machine-renderable terminal state cont
Video understanding has rapidly evolved toward video large language models (VideoLLMs): systems that couple video representations with pretrained large language models and condition generation on a textual prompt. Their strong performance on captioning, question answering, retrieval and temporal grounding comes at a computation and memory cost that grows with frame count and context length, limiti
Before an LLM agent tackles tasks in a new environment, it can inspect available corpora and tools and construct reusable resources such as indices, scripts, or procedural guidance. Most automated adaptation methods, however, rely on task examples, trajectories, or evaluation feedback to decide what to build. Existing task-agnostic approaches avoid this supervision but commit in advance to a prepa
Reasoning language models have made substantial advances on a variety of complex tasks, yet their capabilities remain overwhelmingly English-centric: models primarily reason in English regardless of the language they are prompted in. This is inaccessible for non-English-speaking users, risks losing the intent of the original question, and forgoes knowledge more readily expressed in the target lang
We study how model post-training and test-time inference design affect natural-language proof generation for hard olympiad mathematics. Starting from Nemotron 3 Ultra, we train two specialist checkpoints using supervised fine-tuning and reinforcement learning, and evaluate checkpoint choice, verification, and refinement. Based on these findings, we present an open-model test-time-compute pipeline.
Multimodal entity linking grounds entity mentions in text and images to knowledge-base entries. These systems degrade on rare entities, but prior work measures rarity primarily through popularity-based metrics such as pageviews. We broaden this view using knowledge-graph structural metrics that capture how well an entity is documented and connected. These metrics identify many rare entities that p
Diffusion large language models (dLLMs) achieve high decoding efficiency through block-parallel, arbitrary-order generation, making them attractive for latency-sensitive applications. GUI agents represent a natural testbed for this paradigm, as they must repeatedly perceive screen states and emit structured, spatially grounded actions in real time. However, whether dLLMs can be extended into capab
Existing bias auditing methods typically rely on model outputs, requiring costly benchmarks or judge models and potentially missing internal shifts that never appear in generated text. We propose a reference-based method that audits bias in hidden-state representations across related model variants, for example before and after fine-tuning. Because fine-tuning reshapes representation geometry, abs
We introduce NCP-ArchPreview, a latent-space language model that pushes autoregressive pretraining beyond standard next-token prediction (NTP). Alongside NTP, the model learns through Next Concept Prediction (NCP) to predict discrete concepts that span multiple tokens, introducing an explicit and more challenging concept-level objective while preserving standard token-level autoregressive generati
GUI agents accumulate high-resolution screenshots as the trajectory unfolds, increasing inference latency and memory usage. Training-free visual token pruning can reduce this cost, but cache reuse introduces a fundamental constraint. Once tokens are discarded, the corresponding visual evidence cannot be recovered without re-encoding. Pruning therefore becomes an irreversible admission decision tha
Existing emotional support conversation systems mainly focus on one-on-one seeker-supporter interactions and individual emotional states, leaving interpersonal relations in multi-party scenarios underexplored. In this work, we introduce relation-aware emotional support conversation, a new task that evaluates whether LLMs can capture and utilize the evolving dynamics of relationships to offer more
Large language models (LLMs) have demonstrated remarkable capabilities in reasoning and code generation, raising the prospect that they could assist in developing and optimizing the very infrastructure that powers them. However, existing benchmarks mainly focus on isolated kernels, predefined operators, or pre-specified optimization targets, and therefore fail to evaluate the ability of LLMs to pe
Non-invasive speech decoding remains constrained by the low signal-to-noise ratio of neural recordings, which makes fine-grained reconstruction of phonemes or individual words difficult. Motivated by neuroscientific evidence that high-level semantic representations are distributed across cortical regions and evolve over slower temporal scales, we hypothesize that semantic content may provide a mor
Recent video world models generate increasingly realistic and interactive visual experiences, yet lack reliable mechanisms for maintaining persistent world state and enforcing programmable rules over extended interactions. We introduce Programmable World Model, a framework that decouples world-state evolution from visual observation generation. An agent translates natural-language instructions int
Large language models are unreliable at arithmetic, which is a problem for clinical calculators where a single numerical error changes the recommendation. The standard response is to hardcode each calculator as a validated function, one at a time. We test an alternative: the model does not calculate. Instead, it writes case-specific Python that a restricted local executor runs as a deterministic s
Reacting to sudden physical hazards (catching a slipping plate, dodging a falling knife) is both a meaningful test of embodied intelligence and a hard requirement for deploying multimodal large language models (MLLMs) as the decision coreof household robots. Existing evaluations, however, probe intuitive physics passively through question answering over videos, or target deliberate, long-horizon t
A research idea may be novel, coherent, and scientifically plausible, yet its proposed method may remain insufficiently specified for faithful implementation. We study the codification readiness of implementation-facing research-method specifications, defined by whether they provide sufficient methodological information for a competent implementer or coding agent to construct the intended method w
For research and development (R&D) organizations, the promise of agentic AI is not a better one-time answer. It is a new way to explore complex scientific and engineering problems: pursuing multiple hypotheses, validating them against evidence, learning from what does not work, and adapting their approach as new information becomes available. The post Beyond the benchmark: How an adaptive approach
435 points, 460 comments
Pathway's Baby Dragon Hatchling (BDH) is a brain-inspired, post-transformer architecture that reasons in latent space instead of emitting chain-of-thought tokens. See how Pathway develops and scales BDH on Amazon SageMaker HyperPod, and how BDH-CQ set a new cost-efficiency mark on the ARC-AGI-1 benchmark.
Pathway's Baby Dragon Hatchling (BDH) is a brain-inspired, post-transformer architecture that reasons in latent space instead of emitting chain-of-thought tokens. See how Pathway develops and scales BDH on Amazon SageMaker HyperPod, and how BDH-CQ set a new cost-efficiency mark on the ARC-AGI-1 benchmark.
Amazon SageMaker Feature Store now supports feature-level writes. With the new UpdateRecord API, you can update one or more feature values in a single call without reading or rewriting the entire record. It is available for both the Standard (Amazon DynamoDB) and In-Memory (Amazon ElastiCache) online store tiers.
Governing models across accounts is the next step after automatic model registration. This post extends managed MLflow and Amazon SageMaker AI Model Registry sync to two cross-account governance topologies: a hub-and-spoke pattern that centralizes governance with AWS RAM, and a hybrid pattern that keeps development accounts isolated.
Managed MLflow on Amazon SageMaker AI now syncs richer model metadata (training metrics, evaluation results, inference specs, and lineage) into the SageMaker AI Model Registry, with lifecycle stage promotion. Part 1 shows how to govern candidate models in a single account using IAM guardrails.
See how an MIT researcher uses GPT-5.6 Sol with Codex to autonomously run quantum computing experiments, analyze results, and calibrate qubits.
Wire Amazon Bedrock AgentCore Evaluations into a GitHub Actions pipeline: deploy an AI agent and an OAuth-protected MCP server to AgentCore runtime, invoke the agent with test prompts, score the responses, and automatically block pull requests when agent behavior regresses.
Benchmark two 30B Mixture-of-Experts models, Qwen3-Coder-30B and NVIDIA Nemotron-3-Nano-30B, across G5, G6, G6e, and G7 GPU instances on Amazon SageMaker AI. Compare throughput, latency, and cost-per-token, and see how G7's NVIDIA Blackwell GPUs deliver measurable price-performance gains for real-time LLM inference.
HPE Zerto built an agentic troubleshooting system powered by Amazon Bedrock that runs on-premises inside the customer environment. This post describes the multi-agent architecture, the on-premises deployment model built with Strands Agents, and the engineering challenges of grounding agents in live disaster recovery data.
DiDi built a transparent, self-owned contact center quality assurance (QA) system on Amazon Bedrock, replacing an opaque third-party tool. Intent verification accuracy rose from 38% to 86%, compliance scoring topped 90%, and Voice of Customer trend analysis dropped from hours to minutes across Spanish and Portuguese support.
The post Our Investment in Cylake appeared first on Lightspeed Venture Partners .
<div class="block-paragraph_advanced"><p><span style="vertical-align: baseline;">When building consumer-facing generative AI applications, balancing high generation quality with fast response times across diverse media types, can be challenging. KDDI, a major telecommunications carrier in Japan, tackled this challenge head-on when they developed Buffmee, their consumer Retrieval-Augmented Generati
532 points, 118 comments
Official technical announcement and publication from Hugging Face covering Safety for Whom? Refusing the Right Subset of a Topic, Not the Whole Topic.
665 points, 219 comments
AlphaGenome Atlas maps the molecular effects of 9 billion single-letter DNA variants across the human genome.
374 points, 64 comments
Explore how more capable, affordable AI can expand the work people and businesses can accomplishโand make growth more economical.
Official ElevenLabs technical update and publication covering Ethan Tandowsky joins ElevenLabs as Chief Financial Officer.
Official CMU (Carnegie Mellon AI) technical update and publication covering Season Three of 'Does Compute' Now Available.
ChatGPT Images 2.5 helps turn your ideas, sketches, and reference photos into more personalized, polished images that better reflect your ideas.
Weโre sharing an AI-generated solution to the NavierโStokes Millennium Prize Problem, including a writeup and a formal proof in Lean.
Apply now for OpenAIโs $5 million grant program supporting independent research on how generative AI affects teen development, well-being, and safety.
Mistral raises โฌ3B to make sovereign, open-weight AI the technology frontier
The post From First Check to the Global Stage: Our Continued Partnership with Mistral appeared first on Lightspeed Venture Partners .
Full-duplex speech models require training data that preserves turn-taking, overlap, interruption, and backchannel behavior, yet these signals are entangled across speakers in noisy real-world recordings. We present Conversational Voice, a pipeline that converts real two-speaker excerpts into three complementary training-data artifacts. (1) Separation recovers speaker-specific tracks with stable s
Official Harvey AI technical update and publication covering How Legal Teams Use AI Tools for Contract Generation and Automation.
Official Stanford (HAI) technical update and publication covering AI Legal Review Says Millions Live Under Discriminatory Local Laws.
Delegate File Search and Organization to Harvey
OpenAI is expanding support for journalism with tools, training, and partnerships for students, educators, journalists, and news organizations.
Official Harvey AI release and benchmark update covering Delegate File Search and Organization to Harvey.
Official Mistral AI technical update and publication covering Mistral raises โฌ3B to make sovereign, open-weight AI the technology frontier.
Official Harvey AI technical update and publication covering Post-Training RLM Agents for End-to-End M&A Diligence.
Cognition has raised over $2B at a $48B valuation, led by Andreessen Horowitz and Accel, to build the future of software engineering.
Official Harvey AI technical update and publication covering How Legal Teams Use AI Tools for Contract Generation and Automation.
Inside the megakernel serving engine for North Mini Code
by Julia Andre
Official Harvey AI technical update and publication covering Post-Training RLM Agents for End-to-End M&A Diligence.
Engineers at 1Password use Codex to rapidly build new features and internal tools, reaching production-readiness while maintaining rigorous security policies.
Training capable cyber agents is often treated primarily as a problem of model scale, yet open-weight post-training is constrained more directly by the cost of executable environments, reliable multi-turn supervision, and access to strong teachers. We present a data-centric framework that addresses these bottlenecks through five complementary systems: Choulea analyzes hidden reasoning signatures,
Weak-to-strong generalization asks whether stronger models can learn from weaker supervisors and surpass them. This question is particularly important for successive model generations and multi-domain consolidation, where repeating frontier-scale post-training from scratch can be prohibitively expensive. Yet conventional distillation treats the weak teacher as an optimization target, potentially i
Large Language Models demonstrate remarkable proficiency in static reasoning, yet training them as autonomous agents through Reinforcement Learning (RL) for long-horizon tasks is often hindered by severe reward sparsity. While conventional agent-side warming up via supervised fine-tuning (SFT) can alleviate this, it is frequently limited by data scarcity and constrained exploration. To address thi
Chunk-autoregressive video world models typically condition each generated chunk on one action. An action received during sampling must therefore wait for the next chunk, condition future solver evaluations on a state produced under the previous action, or trigger rollback that repeats completed evaluations. We introduce ActionSplice, an inference framework that formulates this problem as Counterf
We study the problem of navigating cluttered indoor environments with a humanoid robot. Unlike conventional methods that model navigation as a 2D path planning problem, humanoid traversal in cluttered environments requires continuous geometry-aware whole-body adaptation, including coordinated arm placement, torso adjustment, and gait modulation for collision-free movement through complex 3D spaces
SWE-Bench Pro has emerged as a standard benchmark for evaluating software engineering agents on challenging repository-level tasks. However, our analysis work show that its evaluation is undermined by two sources of unreliability: reward hacking, enabled by leakage of gold solutions or hidden evaluation information, and task quality issues, including misleading problem statements and improperly sc
Monocular depth estimation is a ubiquitous yet highly ill-posed computer vision task, with downstream applications in scene reconstruction, computational photography, and robotics, among others. Despite the field's maturity, recent models still struggle to generalize to out-of-distribution inputs and to produce sharp and detailed depth maps. In this paper, we revisit Marigold, a set of techniques
Leading benchmarks for formal theorem proving with large language models are small collections drawn from competition math, such as the IMO and Putnam, that poorly represent field-specific applications. We introduce StochBench, a Lean 4 benchmark of 450 graduate stochastic-processes problems at varying abstraction levels, each paired with its natural-language source. Addressing a field underrepres
We consider the problem of support recovery for sparse binary signals from noisy linear measurements. For sparse Gaussian measurement matrices we identify sufficient conditions on the minimal sample size for maximum-likelihood recovery in the high-SNR regime ds/p to infty, where p denotes the signal dimension, s the number of non-zero components of the signal, and d the expected number of non-zero
Accurate estimation of clinically meaningful gait parameters from monocular video is important for scalable mobility assessment, yet progress is limited by the small scale, restricted viewpoints, and limited visual diversity of existing datasets. We introduce SynthGait-19k, a physically grounded synthetic video dataset containing 19,272 walking videos derived from 6,427 MoCap sequences across 437
We present Miles v0.1, a full-stack, production-ready system for frontier post-training. Building upon the clean design of slime, Miles designs each stage of the reinforcement-learning (RL) training loop around a single principle: components should be verified, clean, and customizable. With accuracy, efficiency, reliability, and scalability as first-class goals, Miles aims to make frontier-scale R
6542 articles sourced historically ยท 100 per page