Nace AI has released Drex for Search, which brings its Drex decision models to document search. Drex for Search is a retrieval layer that helps people and agents find evidence. Coding agents such as Claude Code, Codex, Cursor and OpenCode call it through CLI, SDKs, MCP or REST. In Nace’s evaluation, it scores 92.05% on BrowseComp-Plus with GPT-5, using 12.6 search calls per question. TL;DR Size: a hosted search system built to search over 1M+ files. It pairs with an answering model such as GPT-5, Gemini 3.5 Flash or Qwen 27B. Runs on: Nace’s managed cloud, a dedicated cloud, your own AWS or Azure account, or on-prem and air-gapped. Performance: ranked #1 or #2 on 3 of 4 public leaderboards, per Nace. Best: 77.5% R@1 on VisDocAgentBench, 10 points above Claude Opus 5 (67.5%). What did Nace AI release? Nace AI frames search as a series of decisions. Drex for Search applies its Drex models to those decisions. Drex scores given options with a calibrated probability each, in 1 forward pass. That makes it cheap enough for every agent step. The release targets classic RAG, which Nace sums up as chunk, embed, top-k, hope. The pipeline breaks in 3 cases: the answer sits in a table on page 23 the answer needs 2 hops across files the answer is inside a scanned PDF How does Drex for Search work inside an agent? Nace describes 3 steps: Classify: Before any search runs, Drex decides if the question needs a quick lookup or an agent. The agent route reads documents and queries tables. Squeeze: Drex walks the workspace level by level: folder, file, section, page. Inside a sheet, it narrows to table and column. Each level keeps only what can hold the answer. Answer: Only the kept pages and tables reach the LLM. The answer cites the file and page. Depth is fixed, so Nace says a million files take as many steps as a thousand. How much does it help coding agents? Nace AI tested 9 agent and model pairs on OfficeQA Pro, a 133-question benchmark over U.S. Treasury Bulletins. Key highlights: 6% more answers right, 62% fewer steps, 56% fewer tokens, and 2.3x faster. What does a workspace include? Files: PDFs and scans, xlsx, csv and Parquet, mp3 and mp4, docx, pptx, md, json and txt. Citations: every answer names the file and the page, row or minute it came from. Workspaces: isolated, with their own index and knowledge graph, and shareable with invited members. Storage: upload files or folders, or keep workspaces in your own S3 bucket or Azure Blob container. Zero retention: available with your own storage. Domains: Tax, M&A, SEC 10-K, Contracts, or your own. How does it perform on BrowseComp-Plus? BrowseComp-Plus asks a deep-research agent 830 multi-hop questions over about 100K fixed web documents. With GPT-5, Nace AI reports: 92.05% accuracy, #2 of 86 entries on the leaderboard. 12.61 search calls per question, 16.6x fewer than the only entry ahead (209.27 calls). 87.32% in a single pass with 8.93 search calls, the best of any entry under 10 calls. +20.4 points over the Qwen3-Embed-8B baseline, with the same GPT-5 model. window.addEventListener("message",function(e){if(e.data&&e.data.mtpEmbed==="mtp-drexsearch-a2"){var f=document.getElementById("mtp-drexsearch-a2-frame");if(f)f.style.height=e.data.h+"px";}}); How does it score on document benchmarks? On VisDocAgentBench, an agent finds 1 target page among 2,375 arXiv pages. Drex for Search scores 77.5% R@1 with an open-weight Qwen 27B planner. The best published agent, Claude Opus 5, scores 67.5%. R@1 is 97.5% on direct questions, 77.5% on one bridge, and 57.5% on two. On MADQA, agents answer over 800 PDFs and cite pages. Nace reports 94.7% accuracy with Gemini 3.5 Flash and a page-attribution F1 of 88.2. Its effort calibration (Kuiper) is 6.66. Nace ranks it #1, based on the leaderboard as read on October 20. With open-weight Qwen 27B, Nace reports 93.4%. On EnterpriseRAG-Bench, with 500,000+ internal company documents, Nace reports an overall score of 88.22, which it ranks #3. That is 27.2 points above OpenAI File Search (61.03). How does it compare with the best published results? Ranks and scores are as published by Nace AI, from leaderboards read October 20, 2026. How can developers get access? Drex for Search runs in the Nace API Console, with 4 ways in on one key: CLI: nace-cli on PyPI. SDKs: nace-sdk for Python and nace-sdk for TypeScript. MCP: nace-mcp, for Claude Code, Codex, Cursor or OpenCode. REST: an OpenAPI spec. Key Takeaways Drex for Search gives agents cited pages and tables, not top-k chunks. 92.05% on BrowseComp-Plus with 16.6x fewer search calls than the only entry ahead. 77.5% R@1 on VisDocAgentBench, 10 points above Claude Opus 5. On OfficeQA Pro, agent turns fell for all 9 harness and model pairs Nace tested. Check out the Drex for Search page, API Console, Docs and Hugging Face. Thanks to the Nace AI team for the thought leadership / resources for this article. This article is sponsored by Nace AI. The post Nace AI Releases Drex for Search for AI Agents: 92.05% on BrowseComp-Plus With 12.6 Search Calls appeared first on MarkTechPost.
High-frequency traders like Jane Street and Hudson River Trading are now racing each other for AI computing power. The article Wall Street's fastest traders are building data centers and backing chip startups to stay ahead in AI appeared first on The Decoder.
Toyota wants to turn itself into a global robotics company, and it's betting on humanoid robots and factory automation to get there. The article Toyota is betting $6.4 billion a year on humanoid robots and factory automation appeared first on The Decoder.
Stanford and UC Berkeley researchers released TokenGate, a module for long-context prefill. TokenGate learned token pruning lets an existing LLM skim its input. Only the tokens that matter get full processing. The research team reports 1.4x to 11.3x faster time-to-first-token (TTFT) with no accuracy loss. Peak prefill memory drops 1.5x to 3.4x. Code and checkpoints are on GitHub under Apache-2.0. TL;DR MLP gate: A small MLP gate and LoRA (r=16) added to Qwen3-1.7B or Qwen3.5-2B. Tested on contexts up to 128k tokens. Runs on: Paper timings use 1 A100 at batch size 1. Training used 4 A100s. Performance: Pareto-optimal on 5 tasks against 6 baselines. Best: NIAH 16k at 11.3x faster TTFT and 3.42x lower memory. TTFT gain hits 13.0x at 128k. Worst: Speedup drops to 1.4x on some tasks. What is TokenGate? TokenGate is a drop-in component that adds trainable token pruning to pretrained LLMs. Transformers push every token through every layer. Attention cost grows quadratically with length. Training-free pruning is cheap but loses accuracy at high pruning rates. Architectural changes often need pre-training from scratch. How does TokenGate learned token pruning work? The forward pass splits into 3 phases: Standard forward: layers below gate layer L process all N tokens as usual. Gate module: an MLP scores each token from its post-attention residual at layer L. A top-K selector keeps the highest-scoring tokens. Pruned forward: layers above L process only those K tokens and never create KV entries for pruned ones. Only the document part of the prompt is prunable. The question is never pruned. What is token-gated attention? Top-K selection is not differentiable, which blocks end-to-end training. TokenGate borrows the Mixture-of-Experts router trick. Each kept token’s score becomes a bias on attention logits above the gate. That bias gives gradients a path back to the scorer. A learnable per-head gain and a learnable temperature refine this. The temperature corrects for each query seeing fewer keys. Training uses plain next-token loss. Each step averages a full-context loss and a random-budget loss. So 1 trained gate works across budgets. window.addEventListener("message",function(e){if(e.data&&e.data.mtpH){var f=document.getElementById("mtp-tg-frame");if(f&&e.source===f.contentWindow){f.style.height=e.data.mtpH+"px";}}}); #mtp-tg-wrap hr,#mtp-tg-wrap p:empty,#mtp-tg-wrap del,#mtp-tg-wrap s{display:none!important} How does TokenGate perform on benchmarks? The research team tested Qwen3-1.7B on RepoBench-P, 2WikiMultihopQA, MultiNews, BioDEX and needle-in-a-haystack (NIAH). Baselines were FastKV, SlimInfer, DASH, GemFilter, FlexPrefill and MInference, on the same LoRA backbone. Versus full prefill: 1.4x to 11.3x faster TTFT and 1.5x to 3.4x lower memory. Versus the next-best baseline: 1.2x to 6.4x faster TTFT and up to 2.1x lower memory. 2WikiMultihopQA: at 95% of full-context accuracy, prefill runs 3.0x faster with 2.7x less memory. Longer contexts: a gate trained at 16k reaches 12.8x at 32k and 13.0x at 64k and 128k. Memory savings stay near 3.43x. The learned selector beats FastKV’s heuristic at the same layer. Even random-K selection scores higher on the TokenGate backbone. The research team credits backbone co-adaptation. Does it work on hybrid-attention models? Yes, Qwen3.5-2B mixes Gated DeltaNet and full-attention layers. On NIAH 32k, TokenGate at layer 3 matched full-prefill accuracy. It ran 3.8x faster than full prefill and 2.1x faster than FastKV. Gating at layers 1 or 2 reached 7.8x, within 6% accuracy. FastKV can only prune attention layers. TokenGate vs FastKV, SlimInfer, GemFilter and DASH Self-reported gains use different setups, so they are not directly comparable. Pruning layers come from TokenGate’s Appendix C.2. How do you deploy TokenGate? The GitHub Repo needs Python 3.10+ and CUDA PyTorch 2.9+. An optional FlashAttention-2 key-bias kernel speeds up dense Qwen3. It costs 1.0x to 1.1x a plain call. Without it, inference falls back to FlexAttention. A {state} placeholder marks the prunable document. generate takes a max_kept_prefill_tokens budget. Checkpoints cover all 5 tasks on Qwen3-1.7B, plus a Qwen3.5-2B NIAH gate. Example: liana313/tokengate-niah-qwen3-1.7b. Training mirrors a TRL/PEFT recipe. Each step is 1.3x slower than plain LoRA. Runs converged faster on 4 of 5 tasks. Mean training time was 1.8 versus 2.2 A100 hours. Key Takeaways TokenGate adds learned token pruning to existing LLMs via a small gate and LoRA. Prefill runs 1.4x to 11.3x faster with 1.5x to 3.4x lower peak memory. Token-gated attention makes top-K selection trainable, MoE-router style. It beats the next-best pruning baseline by up to 6.4x on TTFT. It works on hybrid Qwen3.5-2B, where FastKV cannot prune DeltaNet layers. Check out the Paper, GitHub repo and Hugging Face paper page. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well. Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us The post Researchers from Stanford and UC Berkeley Introduce TokenGate: Token-Gated Attention Makes Long-Context Prefill Up to 11.3x Faster appeared first on MarkTechPost.
Anthropic has launched OSS Scanner, a free, opt-in vulnerability scanner for critical open-source projects. The Anthropic OSS Scanner runs the company’s strongest models, including Claude Mythos, against enrolled codebases. Reports go straight to maintainers without human review. That trade delivers faster findings, but shifts triage work onto project teams. 11 days after the October 8 launch, Anthropic reported 77 critical and 803 high model-rated findings. TL;DR Scale: 114 projects accepted from about 650 applications, as of October 19. Runs on: Anthropic’s infrastructure, free. Maintainers supply a Dockerfile; audits run in an isolated VM with no Internet. Performance: 85 of 97 audited critical and high findings (88%) met Anthropic’s disclosure bar. Best: Only 1 of 97 audited findings was a false positive. wolfSSL says 72 of 74 reports were valid. Worst: No human review before delivery. 11 of 97 audited findings duplicated known issues; maintainers report inflated severity. What is Anthropic OSS Scanner? OSS Scanner is a free application from Anthropic that periodically audits open-source repositories with its strongest models. The Frontier Red Team and Safeguards Research team run it together. The design borrows from Google’s OSS-Fuzz, but swaps fuzzers for language models. The main goal is a review bottleneck. Over 6 months, Anthropic’s models surfaced over 29,000 candidate vulnerabilities. Human reviewers triaged roughly 6,000 of them. Nearly 5,000 more reports went directly to maintainers who asked for everything, validated or not. OSS Scanner turns that ad hoc request into an opt-in program. How does OSS Scanner work? Core maintainers enroll by opening a pull request on anthropics/oss-scanner. The PR adds projects//project.yaml, based on the project template. Required fields: repo, primary_contact, and a Dockerfile path. Optional fields: threat_model, pgp for encrypted reports, auto_ccs, homepage, and disabled. The Dockerfile builds the project with network access. The VM then moves to a network without Internet for the audit. Agents in the pipeline double-check bugs, run root cause analysis and propose patches. Each report contains a self-contained reproducer and an explanation of the bug. Where possible, it includes a bisection to the commit that introduced it. A candidate patch is attached when available. An optional threat_model.md defines adversarial inputs, out-of-scope code and a severity rubric. Without it, the scanner guesses. How accurate are OSS Scanner reports? Anthropic team asked its penetration testers to check 97 critical and high findings across 48 projects. 85 (88%) met the bar for Anthropic’s coordinated vulnerability disclosure (CVD) process. 11 were real but duplicated known issues or other scan findings. 1 was invalid. Maintainer data points in the same direction. wolfSSL received 74 reports; all but 2 were valid, and 5 became CVEs. curl patched CVE-2026-92392, a High-severity HTTP/2 upload use-after-free. It affects curl 8.12.0 to 8.22.0 and is fixed in 8.23.0, released October 14, 2026. Anthropic links the timing to benchmark progress. On CyberGym, LLMs went from finding under 20% of vulnerabilities to over 85%. Its broader CVD dashboard lists 6,157 disclosed vulnerabilities across 591 projects, as of October 2. 516 of those are patched. #mtp-oss-scanner-wrap iframe{width:100%!important;border:0!important;display:block!important;background:transparent!important;}#mtp-oss-scanner-wrap hr,#mtp-oss-scanner-wrap p:empty,#mtp-oss-scanner-wrap del,#mtp-oss-scanner-wrap s{display:none!important;}@media (max-width:640px){#mtp-oss-scanner-wrap{margin:16px auto!important;}} (function(){var f=document.getElementById('mtp-oss-scanner-frame');window.addEventListener('message',function(e){if(e.data&&e.data.mtpFrame==='mtp-oss-scanner'&&e.source===f.contentWindow){f.style.height=e.data.h+'px';}});})(); OSS Scanner vs other open-source security programs Sources: OSS-Fuzz disclosure guidelines, OpenAI Codex Security, DeepMind CodeMender. Key Takeaways OSS Scanner gives critical open-source projects free Claude Mythos audits. Reports skip human review; 88% of 97 audited critical/high findings were valid. First 11 days: 77 critical and 803 high model-rated findings. Enrollment is a PR with a Dockerfile; audits run offline. No 90-day deadline applies to unvalidated findings. Check out the launch post, the OSS Scanner FAQ and the GitHub repo. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well. Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us The post Anthropic Launches OSS Scanner, a Free AI Vulnerability Scanner appeared first on MarkTechPost.
Nvidia's new data engine Flora labels one hour of robot video for 72 cents instead of $88 with GPT-6 Astra, while scoring slightly higher on quality. The open system, built on three fine-tuned Qwen3.5 models, catches errors in robot actions far more reliably than any competitor but sometimes falls short on descriptive accuracy. Nvidia is positioning Flora as a key piece of its strategy to control the entire robotics value chain. The article Nvidia open-sources Flora to make robot video labeling 125x cheaper than GPT-6 Astra appeared first on The Decoder.
A new benchmark by P-Zero Research measures how well AI models design and evaluate experiments on their own. The results are improving so fast that Anthropic co-founder Jack Clark now thinks self-improving AI could arrive by 2027. The article Anthropic co-founder Jack Clark moves his prediction for self-improving AI forward to 2027 appeared first on The Decoder.
