October 3 – 9, 2026
Saturday to Friday, Brasília time.
- 3,306pieces
- 836sources
- 33new models
- 6exploited vulnerabilities
- 61major incidents
The Week's Top Stories
What mattered most, each in its maker's or first source's words.
-
1
Claude Haiku 5.5 launch
Claude Haiku 5.5 launchWe just launched Claude Haiku 5.5, the cheapest, fastest, and most capable small model we’ve released, designed for high-volume, cost-sensitive tasks. For more information, see our blog post: Claud…
Also reported by Claude API Release Notes · Vercel · AWS Machine Learning · GitHub Changelog · What's New with AWS · Netlify
-
2
EmbeddingGemma 2: an open, lightweight multimodal embedding model
Also reported by Google · Developers · Google for Developers · Qdrant
-
3
Added Ultrafast mode for GPT-6.1 Sol in the Responses API.
Feature · gpt-6.1-sol · v1/responses. Added Ultrafast mode for GPT-6.1 Sol in the Responses API. Use gpt-6.1-sol with service_tier: "ultrafast" to reduce the time between generated output tokens. It is available to all A…
Also reported by Codex Changelog · OpenAI · What's New with AWS
-
4
Gemini Nano Banana 2.1 generally available (GA)
Gemini Nano Banana 2.1 generally available (GA): Released Gemini Nano Banana 2.1 (gemini-nano-banana-2.1), the latest high-efficiency image generation and conversational editing model. An update to Nano Banana 2 (gemini-…
-
5
Introducing Mistral Large 4
Also reported by Vercel · Mozilla.ai
-
6
We've added capabilities.server_tools to the Models API.
We've added capabilities.server_tools to the Models API. GET /v1/models and GET /v1/models/{model_id} now report whether each model accepts the web search and code execution tools. To check whether a model accepts the co…
-
7
The verifiability litmus test for agent design
Whether answers can be verified decides AI agent design. Case studies from agentic search, deep research, RAG and coding.
-
8
We've added capabilities.thinking.types.disabled to the Models API.
We've added capabilities.thinking.types.disabled to the Models API. GET /v1/models and GET /v1/models/{model_id} now report whether each model accepts thinking: {type: "disabled"}, which turns thinking off. See Using the…
Security
Vulnerabilities newly known to be exploited (CISA), then the week's most relevant advisories.
- CVE-2026-88779: Citrix NetScaler Improper Restriction of Operations within the Bounds of a Memory Buffer Vulnerability
Citrix NetScaler ADC (formerly Citrix ADC) and Citrix NetScaler Gateway (formerly Citrix Gateway) contain an improper restriction of operations within the bounds of a memory buffer vulnerability that could allow for a de…
- CVE-2015-5477: ISC BIND Data Processing Errors Vulnerability
ISC BIND contains a data processing errors vulnerability that could allow remote attackers to cause a denial of service via TKEY queries.
- CVE-2016-3081: Apache Struts Command Injection Vulnerability
Apache Struts contains a command injection vulnerability that could allow remote attackers to execute arbitrary code via method:prefix when Dynamic Method Invocation is enabled.
- CVE-2023-22894: Strapi Cleartext Storage of Sensitive Information Vulnerability
Strapi contains a cleartext storage of sensitive information vulnerability that could allow attackers with access to the admin panel to discover sensitive user details via the query filter. The impacted product(s) could…
- CVE-2021-3199: ONLYOFFICE Docs Server Path Traversal Vulnerability
ONLYOFFICE Docs contains a path traversal vulnerability that can occur when JWT is used, via a /.. sequence in an image upload parameter and could allow for remote code execution.
- CVE-2015-3306: ProFTPD Improper Access Control Vulnerability
ProFTPD contains an improper access control vulnerability that could allow remote attackers to read and write to arbitrary files via the site cpfr and site cpto commands.
- Critical: PraisonAI: CodeAgent Executes LLM-Generated Code Without Sandboxing and Leaks All Environment Secrets
Affects praisonaiagents. CVE-2026-61447.
- Why the AI Attack Surface Extends Your Stack
Integrating AI into enterprise applications creates risk across every layer of the stack, not just the model. This post maps the full AI attack surface, shows how a prompt injection can chain into credential theft and cr…
- AI Agent Security: Six Controls From Nine Real Incidents
An AI agent runs with your credentials and acts before you can stop it. Nine incidents show what that costs. Six controls limit the damage: sandbox the agent, scope its credentials, lock its configuration, log every call…
- Critical: Langflow: PythonREPLComponent executes unsandboxed Python code, enabling authenticated RCE and privilege escalation
Affects langflow. CVE-2026-10561.
- Critical: PraisonAI: AICoder Arbitrary File Write and Command Execution via LLM Tool Calls
Affects praisonai. CVE-2026-61445.
Major Incidents
Outages the providers themselves rated major or critical.
- Codecov app and services are down
- Issue with the Start Build dialog in the Website UI
- Increased processing times and errors for Universal-3.5 Pro on the EU async endpoint
- Elevated errors for Claude Mythos 5.1 and Claude Fable 5.1
- Logging into Zapier using Facebook or Microsoft SSO shown a blank page after authentication.
- Incident with Actions
- Elevated errors on Direct Uploads
- Website and services partially unavailable due to high database load
- Increased errors on Temporal Cloud control plane, Cloud Ops API, and Temporal UI
- Disruption with some GitHub services
- DeepSeek v4.1 outage on Model APIs. Dedicated Inference is not affected
- EU Agents disruption in SIP calls
- Elevated errors for Claude Opus 5.5
- Increased Processing Times and Errors on US Endpoint for Universal-3.5 Pro Async model
- Delayed Serverless Metrics in Azure UAE North region
- INC20000263
- Issues with Pulumi OIDC (inbound)
- Vercel Auth login failures for pages with Deployment Protection enabled
- Delayed delivery of 10DLC SMS messages to AT&T destinations
- Sentry dashboard down
- Several services are degraded
- Elevated errors loading usage data on platform.claude.com
- Partial Outage: Video Ingest and Delivery
- Elevated errors on platform.claude.com
- Incident with Git Operations, Pull Requests and Actions
- Incident with Git Operations, Issues, Actions and Pull Requests
- [SERVERLESS][GCP][us-central1] Newly created indexes failing to initialize
- Database Connection Issues and 504 Errors on a Subset of Sites
- Accessibility DevTools is facing issues
- Spectra (Accessibility Rule Engine) is facing issues
- [Medium] Issue with Enterprise Event Reports
- LHR network issues
- [Critical] Issues with Multiple Box Services
- Elevated listener latency and error rate impacting Studio loading
- Experiencing high error rate in AZURE in Germany West Central region
- JFrog Cloud Platform - Xray Service Disruption in Azure US East
- Google Cloud Storage customers are experiencing elevated latencies and error rates when performing upload or write operations.
- Agent SIP Conversation Disruption
- Increased Dubbing Latency and Failure Rate
- Low Code Automation - Test Creation is facing issues
- Some users may have issues using Canva
- Delayed or Unavailable project operation in GCP us-central1
- 2026-October-08 Service Incident
- Elevated listener latency affecting Studio performance
- Sprite creates failing
- Bandwidth Local Market Impairment - 328 (Dayton, OH)
- Bandwidth Local Market Impairment - 977 - Macomb, IL
- Bandwidth Local Market Impairment - 122 - Manchester, NH
- Delay in data appearing in UI and API
- Experiencing high error rate in AZURE in West Europe region
- Issue impacting Atlas Search availability on certain clusters
- Some ChatGPT content may be unavailable on Android
- Bandwidth Local Market Impairment - 554 - Longview, TX
- Hub Registry Authenticated Actions Failing
- Workspace admins may be unable to manage plugins
- Failing function calls
- We've identified a problem and are working on a fix
- AWS us-east-1 connectivity failures
- Possible outage in AWS us-east-1
- Delays in Drains, Observability, and Workflow in Washington (iad1)
- System Errors in AP-SOUTHEAST-1
Releases
Major and minor versions only — no patches.
- Anthropic Python SDK 1.12.0: api, add claude-haiku-5-5 and typed computer and browser toolset tool calls (f6346e6)
1.12.0 (2026-10-07) Full Changelog: v1.11.0...v1.12.0 Features api: add claude-haiku-5-5 and typed computer and browser toolset tool calls ( f6346e6 ) api: add disabled to thinking types in model capabilities ( 78d815d )…
- Anthropic Python SDK 1.13.0: api, add types for the Chat and Cowork unified analytics metrics
Features api: add types for the Chat and Cowork unified analytics metrics api: add workflows, multiagent configuration and thread status filtering to Managed Agents Chores internal: name the default API version header va…
- sdk-v0.133.0
Features api: add types for the Chat and Cowork unified analytics metrics api: add workflows, multiagent configuration and thread status filtering to Managed Agents Chores internal: remove the next branch rule from CLAUD…
- Xcode 27.1 RC (27A9275)
View downloadsView release notes
- CPython 3.15.0: Python 3.15.0
Python 3.15.0
- v0.39.0
What's Changed Support LynnReal light Minimax-H3 vae by @kijai in #16657 Update embedded docs to v0.5.13 by @comfyui-wiki in #16618 Use higher quality defaults for Save Video encoding. by @comfyanonymous in #16663 feat:…
- chore: release @cypress/webpack-dev-server-v6.1.0
[skip ci]
- Diffusers 0.41.0: QwenImage 2.1 pipeline and more
Tip This release brings Qwen-Image 2.1 to Diffusers, with text-to-image generation, image editing, native transparency, and LoRA training. Starting with this release, we’re adopting the Transformers release philosophy: c…
- electron v44.6.0
Release Notes for v44.6.0 Features Added the experimental protocol.registerSource() to serve a custom scheme from directories on disk without a JavaScript handler. #54522 (Also in 45 ) Fixes Fixed a crash on launch on ma…
- electron v44.7.0
Release Notes for v44.7.0 Features The macOS autoUpdater now streams update downloads to disk, resumes an interrupted download on the next check instead of starting over, and applies a binary delta when the update feed o…
- Express 5.3.0: Bug fixes
Highlights Bug fixes Fixed HTTP header conflict between Content-Length and Transfer-Encoding in res.send - by @YuryShkoda in #4893 Fixed the behavior of res.send() to prevent conflicts between Content-Length and Transfer…
- FastAPI 0.143.0: Automatic OpenTelemetry exporter setup is now opt-in. If you rely on FastAPI to configure exporters from…
Automatic OpenTelemetry exporter setup is now opt-in. If you rely on FastAPI to configure exporters from OTEL_* environment variables, set FASTAPI_OTEL_AUTO_CONFIGURE=true or use FastAPI(telemetry={"auto_configure": True…
- Release v0.63.0
What's Changed fix(cli): display retry progress indicator during connection recovery ( #28340 ) by @amelidev in #29468 Changelog for v0.61.0-preview.1 by @gemini-cli-robot in #29469 Changelog for v0.61.0 by @gemini-cli-r…
- v1.19.0
https://github.com/gleam-lang/gleam/blob/v1.19.0/CHANGELOG.md
- Google Gen AI SDK 2.29.0: Support "allowlist", "disabled" in environment network config. (fa4acc7)
2.29.0 (2026-10-07) Features Support "allowlist": "disabled" in environment network config. ( fa4acc7 ) Update discovery doc ( 84bf19c ) Bug Fixes Only log direct AFC warning when functions are present in the config ( 6e…
- Goose 1.54.0: Prompt for goose mode during first-time configure #12629
Features Prompt for goose mode during first-time configure #12629 HTML session export with an interactive viewer #11977 Bug Fixes Adaptive thinking for Claude Sonnet and Haiku 5.5 #12761 Treat a client extension selectio…
- LanceDB v0.40.0
This stable release uses Lance 13.0.0 and is maintained on release/v0.40. Java packaging: Use com.lancedb:lancedb-core:0.40.0. The Maven artifacts were built from release-branch commit 711deef7, which corrects version se…
- Laravel 13.35.0: [13.x] Allow fake assertions to accept an array of properties by @jackbayliss in #61770
[13.x] Allow fake assertions to accept an array of properties by @jackbayliss in #61770 [13.x] Support percentages for queue:work --memory by @jackbayliss in #61753 [13.x] Let long-running scheduled commands see schedule…
- v1.104.0
Verify Docker Image Signature All LiteLLM Docker images are signed with cosign. Every release is signed with the same key introduced in commit 0112e53. Verify using the pinned commit hash (recommended): A commit hash is…
- operator: v0.12.0
0.12.0 (2026-10-07) BREAKING CHANGES operator: Make schemas field required in storage spec ( #24716 ) Features operator: Add CPU and memory utilization ratio telemetry recording rules ( #23379 ) ( 4cc1c43 ) operator: Add…
- Next.js 16.4.0: Check out the 16.4 announcement post to get an overview of the changes
Check out the 16.4 announcement post to get an overview of the changes. Core Changes Show compiler plugin warning in more situations: #75682 fix(scripts): correct typo in rm.mjs error message: #87015 docs: improve clarit…
- 2026-10-07, Version 26.11.0 (Current), @aduh95
Notable Changes [ 232f178825 ] - (SEMVER-MINOR) buffer: add isLatin1 (James M Snell) #66298 [ 27778036de ] - (SEMVER-MINOR) buffer: add Buffer.stringLength() (Matteo Collina) #66064 [ 4f84c4bda2 ] - build, doc: move to r…
- Nuxt 4.6.0: v4.6.0 is the next minor release
v4.6.0 is the next minor release. Some news Nuxt CLI v4 Alongside the release of Nuxt v4.6, today also brings a new major release of the Nuxt CLI: @nuxt/cli v4. It ships as a dependency of nuxt, so you'll get it automati…
- Nx 23.3.0: angular, support ngrx v22 (#36950)
23.3.0 (2026-10-07) Features angular: support ngrx v22 ( #36950 ) angular-rspack: scaffold and run SSR apps on the @angular/ssr application engine ( #36298 ) core: emit a runbook and per-step instructions for orchestrate…
- ONNX Runtime CUDA Plugin EP 0.3.0
This release brings sparse attention and packed continuous-batching support, INT2 and mixed-width QMoE execution, decode-kernel tuning, and reliability fixes to the ONNX Runtime CUDA Plugin EP. Highlights include new QSA…
- ONNX Runtime v1.31.0
ONNX Runtime 1.31.0 stabilizes model-package and EPContext data APIs, improves CPU model loading and quantized MoE inference, expands device-based execution-provider selection, and strengthens model-loading and runtime r…
- v0.40.0
What's Changed Models run on MLX on Apple Silicon by default In this release, on Apple Silicon devices, model architectures supported by the MLX runtime will automatically run on MLX. ollama pull qwen3.8 ollama run qwen3…
- v0.40.0
What's Changed Models run on MLX on Apple Silicon by default In this release, on Apple Silicon devices, model architectures supported by the MLX runtime will automatically run on MLX. ollama pull qwen3.8 ollama run qwen3…
- v0.40.0
What's Changed Models run on MLX on Apple Silicon by default In this release, on Apple Silicon devices, model architectures supported by the MLX runtime will automatically run on MLX. ollama pull qwen3.8 ollama run qwen3…
- v0.40.0
What's Changed Models run on MLX on Apple Silicon by default In this release, on Apple Silicon devices, model architectures supported by the MLX runtime will automatically run on MLX. ollama pull qwen3.8 ollama run qwen3…
- Ollama 0.40.0: Models run on MLX on Apple Silicon by default
What's Changed Models run on MLX on Apple Silicon by default In this release, on Apple Silicon devices, model architectures supported by the MLX runtime will automatically run on MLX. ollama pull qwen3.8 ollama run qwen3…
- OpenAI Codex 0.161.0: New Features
New Features GPT-6.1 Sol is now the default model in the bundled and Amazon Bedrock catalogs. ( #49318, #49339 ) Amazon Bedrock supports multi-agent V2 and Ultra reasoning on compatible models; Bedrock Mantle also accept…
- OpenAI Codex 0.162.0: New Features
New Features Add tools for creating and listing managed Git worktrees from trusted local projects when the worktrees feature is enabled. ( #50148 ) Pin tasks in the agent Command Center with p and keep them in a shared P…
- v7.28.0
7.28.0 (2026-10-02) Features api: add custom voice creation and agent session events ( #2858 ) ( 11b9283 )
- OpenAI Node SDK 7.29.0: agents, observe local tool failures (#2861) (3808241)
7.29.0 (2026-10-05) Features agents: observe local tool failures ( #2861 ) ( 3808241 ) agents: run local tools during session creation streams ( #2862 ) ( 7423ac3 ) tools: support tool search (defer_loading=true) in type…
- OpenAI Node SDK 7.30.0: api, add standalone Decisions support (#2886) (7195644)
7.30.0 (2026-10-06) Features api: add standalone Decisions support ( #2886 ) ( 7195644 )
- OpenAI Node SDK 7.31.0: api, add prewarmed hosted environments (#2898) (7a0be86)
7.31.0 (2026-10-09) Features api: add prewarmed hosted environments ( #2898 ) ( 7a0be86 )
- OpenAI Node SDK 7.32.0: api, add agent environment suspension and expiration (#2904) (6e4ed63)
7.32.0 (2026-10-09) Features api: add agent environment suspension and expiration ( #2904 ) ( 6e4ed63 ) Bug Fixes deps: bump next from 15.5.25 to 15.5.27 in /ecosystem-tests/vercel-edge ( #2896 ) ( 88b4c05 ) runners: pre…
- OpenAI Python SDK 3.25.0: agents, observe local tool failures (#4028) (95d9645)
3.25.0 (2026-10-06) Features agents: observe local tool failures ( #4028 ) ( 95d9645 ) agents: run local tools during streamed session creation ( #4027 ) ( 919b623 ) api: add agent turn items and usage source grouping (…
- OpenAI Python SDK 3.26.0: api, add standalone Decisions support (#4033) (b99e3e4)
3.26.0 (2026-10-06) Features api: add standalone Decisions support ( #4033 ) ( b99e3e4 )
- OpenAI Python SDK 3.27.0: api, add prewarmed hosted environments (#4043) (c511a77)
3.27.0 (2026-10-09) Features api: add prewarmed hosted environments ( #4043 ) ( c511a77 )
- OpenAI Python SDK 3.28.0: api, add agent environment suspension and expiration (#4046) (6aa25a9)
3.28.0 (2026-10-09) Features api: add agent environment suspension and expiration ( #4046 ) ( 6aa25a9 ) Chores deps: bump astral-sh/setup-uv from 10.1.0 to 10.2.0 ( #4026 ) ( 8bfffd4 ) deps: bump CodeQL init and analyze…
- oxlint v1.87.0
Features 42dcfd2 linter/react/no-unescaped-entities: Implement suggestion ( #27292 ) (Mikhail Baev) cd4510d linter/unicorn/no-useless-switch-case: Implement suggestion ( #27245 ) (Mikhail Baev) 1a7c902 linter/react/jsx-n…
- oxfmt v0.72.0
BREAKING CHANGES e2c68b1 oxfmt: [ BREAKING ] Format parser:markdown files by oxc_formatter_markdown ( #27256 ) (leaysgur) Features b538e3c formatter_css: Parse embedded CSS as a block's contents ( #27284 ) (leaysgur) Bug…
- Playwright 1.64.0: New page.webmcp and frame.webmcp give access to the tools that a page registers through the experimental…
WebMCP New page.webmcp and frame.webmcp give access to the tools that a page registers through the experimental WebMCP browser API, so you can test them like any other part of your app: const browser = await chromium. la…
- sdk/v3.268.0: pin maven to latest 3.9.* version (#25082)
Maven 3.10 has a bug, where it ignores -Dorg.slf4j.simpleLogger.logFile=System.err, and thus writes everything to stdout. This breaks pulumi-java plugin downloads, as the JSON is now mixed with the regular output and we…
- Pulumi 3.268.0: [engine] Negotiate OutputValues on Invoke so SDKs can pass dependency-carrying values in args and receive…
3.268.0 (2026-10-07) Features [engine] Negotiate OutputValues on Invoke so SDKs can pass dependency-carrying values in args and receive them in the response, and providers can process them, gated by a new accepts_outputs…
- pkg/v3.268.0: Changelog and go.mod updates for v3.268.0 (#25095)
Co-authored-by: github-actions github-actions@github.com Co-authored-by: Thomas Gummerer t.gummerer@gmail.com Co-authored-by: Claude Opus 5.5 (1M context) noreply@anthropic.com
- core-v2.49.0
pydantic-core release v2.49.0
- v2.14.0 2026-10-08
What's Changed The highlights of the v2.14 release are available in the blog post. Several minor changes (considered non-breaking changes according to our versioning policy ) are also included in this release. Make sure…
- core-v2.50.0
pydantic-core release 2.50.0
- Qdrant 1.19.0: #9669, #9684, #9950 - Unify definition of memory usage strategy for collection components. Use "memory"…
Change log Features https://github.com/qdrant/qdrant/milestone/50 - TurboQuant 4-bit as a datatype of primary vector storage. Only store 4-bit quantized vectors and spare disk space on original vectors. [ docs ] #9669, #…
- v5.0.0-0
5.0.0-0
- Ruff 0.17.0: Released on 2026-10-09
Release Notes Released on 2026-10-09. The executables in our macOS and Windows release archives and ruff wheels are now code-signed. macOS executables are signed with an Apple Developer ID certificate and notarized by Ap…
- swc_core@v82.0.0
chore: Publish crates with swc_core v82.0.0
- v2.120.0
Supabase CLI v2.120.0 — 2026-10-06 A mix of usability improvements and bug fixes, with two major additions for debugging and observability: service logs now persist to disk and ship to Analytics, and CLI runs can export…
- TRL 1.15.0: Fused LM head, up to 6.9× longer sequences on the same GPU
Features Fused LM head: up to 6.9× longer sequences on the same GPU SFT, DPO, KTO, GRPO, RLOO and Distillation now score tokens with a fused LM head: a Triton kernel projects the hidden states through the LM head in tile…
- Release v5.19.0
Release v5.19.0 New Model additions EmbeddingGemma2 EmbeddingGemma 2 is a multimodal embedding model from Google built on the Gemma 4 architecture. It encodes text, images, audio, and video, individually or combined in o…
- gfx950-tutorial-v3.0
gfx950 tutorial pin v3.0: upstream main at 3b0c7f0 (PR #12209, the …
- 1.141.0
https://code.visualstudio.com/updates/v1_141
- v1.40.0 - Alter Schema Drop Vector Index, HFresh MUVERA, RQ-4 quantization
Breaking Changes none Namespaces (GA) Namespaces add control-plane and data isolation between users on a shared cluster. Namespaces: Gate tenant actions by @dirkkul in #12513 Namespaces: Add plumbing for namespace existe…
- v0.6.0
Overview llama.cpp v0.6.0 introduces the new llama_batch_ext extended batch API (with llama_process ) for mixed token/embedding inputs and MTP/deepstack state embeddings, adds support for the GLM-5.3-Flash (GLM5-Next) 32…
- aqua-registry-v2026.10.0
Release aqua-registry 2026.10.0
- mise-interactive-config-v2026.10.0
Release mise-interactive-config 2026.10.0
- openai 7.28.0
The official TypeScript library for the OpenAI API
- next 16.4.0
The React Framework
- openai 7.30.0
The official TypeScript library for the OpenAI API
- @anthropic-ai/sdk 0.132.0
The official TypeScript library for the Anthropic API
- @google/genai 2.28.0
[](https://www.npmjs.com/package/@google/genai) [](https://www.npmjs.com/package/@google/ge…
- openai 7.31.0
The official TypeScript library for the OpenAI API
- @anthropic-ai/sdk 0.133.0
The official TypeScript library for the Anthropic API
- openai 7.32.0
The official TypeScript library for the OpenAI API
- pnpm 12.10.0
This release adds an experimental loaded node linker, lets pnpm-lock.yaml record resolution settings, and reads cached registry metadata faster. It also carries several security fixes, including one that stops a dependen…
- pnpm 12.11.0
This release adds Rust toolchain management, links the agent skills that dependencies ship, adds the permissions setting, and keeps the colors of streamed script output. Minor Changes pnpm install now links the agent ski…
- uv 0.13.0: Released on 2026-10-09
Release Notes Released on 2026-10-09. uv 0.13.0 makes Python 3.15 the default stable Python version. We've also included several breaking changes to improve correctness, performance, and compatibility, described below. W…
- vLLM 0.31.0: This release features 717 commits from 307 contributors (96 new)!
v0.31.0 Highlights This release features 717 commits from 307 contributors (96 new)! DeepSeek-V4.1-Flash performance: FlashMLA mega attention with the V4.1 NVFP4 compressed KV cache is now the SM100 default ( #56935 ); D…
- proto-v0.5.0
Release vllm-proto 0.5.0
Research
The papers most useful to a developer this week.
- One Word Opens the Gate: The Option-Channel Attack on Typed Decision Models as Agent Guardrails
A typed decision model reads a piece of text and returns a probability over caller-defined options, each with a short written definition, generating no text. Recent work places these models in agent systems as guardrails: the component that reads a proposed tool call or incoming…
- Revealing and Mitigating the Impact of LLM-Generated Content on Retrieval-Augmented Generation Systems
In the era of Large Language Models (LLMs), the Internet is flooded with LLM-generated content, including both aligned LLM-generated content that contains equivalent facts to human-written content and distorted LLM-generated content with factual discrepancies. However, there is a…
- ARC: A Reasoning Recipe for Robot Foundation Models
The prevailing approach to improving robot foundation models (RFMs) relies on larger models, more robot demonstrations, and costly training at scale. We show that there exists an effective and efficient complementary approach: the right reasoning recipe can substantially improve…
- Reasoning-Informed Visual Editing
Large Multi-modality Models (LMMs) have made significant progress in visual understanding and generation, but still face challenges in visual editing, particularly in following complex instructions, preserving appearance consistency, and supporting flexible input formats. To stud…
- Beyond Spatio-Temporal Priors: A Generalizable Approach for Dense Correspondence Matching
Dense correspondence matching has historically been bounded by simplifying spatio-temporal priors, such as smooth motion and rigid geometry. While effective for classical tasks, these assumptions break down in image editing and reference-guided generation (IEG), where transformat…
End of the week.← The week before · The week after →