Search AI news

Showing our work: a controlled experiment in context engineering

We ran one AI skill 80 times against four context sources. Engineered summaries cut cost 48.5% and run time 43.2%.

Context Engineering (Agent Memory) for AI Agents Starts with Retrieval

Agents waste time and context on clumsy searches, oversized results and stale data. Learn the retrieval techniques LangChain's Deep Agents harness uses to get the right context in.

Context Engineering with Deep Agents: How To Remember Without Keeping Everything In Context

Long-running agents degrade as context piles up. Learn how offloading, summarization, skills, subagents, and memory keep your agent's context window focused.

Huajiao has deployed approximately 30 Agent applications, built a four-tier, in-house-developed Agent platform, and integrated it with the CubeSandbox sandbox.

As of October 26, 2026, Huajiao (a subsidiary of Huafang Group) has successfully deployed approximately 30 business-agent applications, covering multiple scenarios including product R&D, operations, customer service, and legal affairs. Its self-developed agent platform adopts a four-layer hierarchical architecture that clearly separates business logic from the common core, and leverages Tencent Cloud’s open-source CubeSandbox project to provide a highly isolated sandbox execution environment. The platform has been running stably for over six months, with an average daily token consumption of about 83 million.

Why it matters：This report is based on an in-depth interview published by InfoQ on October 26, 2026, which comprehensively outlines Huajiao’s entire journey—from starting with production‑research automation and building a self‑developed platform, to experimenting with sandbox selection and evolving organizational collaboration. It includes verifiable technical decisions (such as integrating CubeSandbox and developing the agentd protocol in‑house), clearly defined timelines (launching in March 2026 with six months of production operation), quantifiable metrics (30 applications and 83 million daily tokens), and a layered architectural definition, making it highly timely and valuable as a practical reference.

How Ramp's internal AI agents crossed a million sessions

Ramp shares the agents, sandboxes, and dashboards it built by treating internal AI as analytics engineering.

From mainframe to AI-powered maturity: Verisk's data modernization journey

Verisk modernized decades of insurance data, from mainframe to dbt, to get AI-ready.

How DocuSign cut analytics engineering time in half by pairing Claude Code with dbt v2, dbt MCP, and agent skills

DocuSign wired Claude Code, dbt Agent Skills, dbt MCP, and dbt v2 into one workflow and measured a 50% productivity gain.

From prompt to PR: how Careem’s AI writes and validates dbt models before a human ever sees them

Careem's AI agent writes and validates dbt models end to end before a human ever reviews the PR.

How a data scientist at Kilo built the business context layer his AI agents were missing

How a data scientist built an AI context layer for organizational knowledge a dbt project alone can't hold.

What a swarm of rogue AI agents can teach us about having better ideas

Benn Stancil uses the Hugging Face hack to explain why the most useful ideas come from the weirdest connections.

How to solve dashboard debugging in six minutes with an MCP-connected AI agent

An AI agent used dbt lineage from inside the BI layer to answer three data mysteries in the time it takes to get coffee.

Which Answer Did the Seventeen-Year-Old Write? A Behavioral Case Study of Author-Age Misattribution, Miscalibrated Confidence, and Non-Corrective Self-Explanation in Six Frontier Language Models

arXiv:2610.31946v1 Announce Type: new Abstract: Three students answered the same exam prompt by hand: "What is happiness? Are you happy?" Their ages, recorded at collection, are 17, 28, and 45. The 17-year-old's answer is hedged and formally phrased. Asked to match each answer to an age and gender label, most frontier language models confidently assign it to an adult. This paper reports a structured case study of that failure across six models: Claude Fable 5, Gemini 3.1 Pro, GPT-5.6 Sol, Grok 4.5, GPT-4o (2024), and DeepSeek V4. We ran 1,081 API sessions and scored 3,225 responses across six experimental conditions that manipulate framing, disclosure format, and audience-awareness instruction. Three of six models perform at or near chance (1/3) on the matching task while reporting 73-88%

RSI-Opt: Recursive Self-Improvement Optimizer Agents For Model Training

arXiv:2610.32026v1 Announce Type: new Abstract: Conventional neural-network training typically follows a predefined optimization recipe, motivating a LLM agent that monitors training dynamics and adjusts optimizer settings as training progresses, but a fixed agent policy does not accumulate experience across runs. We introduce RSI-Opt, a plug-and-play Recursive Self-Improvement Optimizer Agents which evolves the knowledge that guides this process: a prompt, a playbook of procedures, a library of tested optimizer code, and a decision cadence. A meta-agent rewrites this policy using evidence from earlier runs, and an archive retains evaluated policies for further evolution. Across 5 generations on three training tasks, the meta-agents develop reusable procedures and code and conduct their ow

CEMIR vs. FISER: a comparative study of reasoning-based and enrichment-based approaches to pragmatic intent recognition

arXiv:2610.32022v1 Announce Type: new Abstract: Intent recognition in human-robot interaction requires agents to resolve ambiguous natural language instructions by inferring the speaker's underlying goals. Research has increasingly focused on two notable paradigms for addressing this task: structured symbolic reasoning and multimodal input enrichment. However, it remains unclear whether enrichment-based methods provide benefit in settings where the ambiguity concerns unstated human intent rather than missing observable scene information. In this paper, we present a comparative and diagnostic study benchmarking FISER - a framework utilizing structured social and embodied reasoning - against CEMIR (Context-Enriched Multimodal Intent Recognition), a method that performs multimodal enrichment

How Much Do You Trust A Dog? Quantifying Hint-Following in Reasoning Language Models

arXiv:2610.32055v1 Announce Type: new Abstract: Large language models (LLMs) often exhibit sycophancy, altering their answers to align with user suggestions even on objective factual and reasoning tasks. Yet the quantitative form of this influence--- and whether it is calibrated to source reliability--- remains poorly understood. In this work, we investigate this behavior through hint-induced answer shifts on MMLU questions, evaluating prompts of the form ``[Source] thinks the answer is (X)''. We systematically vary the source to which the hint is attributed (a professor, a friend, or a dog) across three open-source model families, Qwen, Gemma, and SmolLM. We show that the change induced by the hint follows a predictable structure well-characterized by an exponential tilt: the hint induces

Reproducible Evaluation as the Substrate for Safe Prompt Compression in Production Agents

arXiv:2610.31972v1 Announce Type: new Abstract: Evaluating conversational agents that act on live, mutable enterprise data is hard: replayed sessions drift as backend state changes, and a benchmark built today can silently go stale as the data it describes moves on. We present an offline evaluation harness for a production conversational agent that stays reproducible despite the live data it operates on changing, and that can be regenerated on demand once it goes stale. We then use this harness as the fitness function for GEPA-based prompt optimization under a correctness gate that guards against reward hacking. Applied to a production merchant-facing agent, correctness-gated compression shrinks the agent's full prompt surface by 3.5-7.6% without regressing held-out task correctness, consi

POS-Signature Demonstration Retrieval for In-Context Dependency Parsing

arXiv:2610.31966v1 Announce Type: new Abstract: Demonstrations for in-context learning are often retrieved by semantic similarity, which may be a poor proxy for relevance in structured prediction. We study demonstration selection for Universal Dependencies parsing with a POS-signature policy that represents sentences with weighted part-of-speech statistics and retrieves three length-compatible, gold-parsed examples from the labeled training split. We evaluate two open-weight language models without fine-tuning on all 2,077 sentences of the UD English Web Treebank test set, comparing the policy with shot- and format-matched fixed and semantic three-shot baselines, as well as zero-shot, chain-of-thought, and critique-refinement systems. POS-signature retrieval yields the highest labeled atta

Semantically Rich, Metrically Affect-Blind: A Geometric Blind Spot in Modern Text Embedders

arXiv:2610.32122v1 Announce Type: new Abstract: We report a systematic dissociation in modern text embedders: sentiment is almost perfectly decodable from their embeddings, yet almost entirely absent from their metric. Across eleven frozen encoders spanning 2021 to 2026, including 2026-generation models and the 4B-parameter LLM embedder Qwen3-Embedding-4B, a linear probe recovers sentence polarity at up to 91.3% accuracy while cosine similarity separates polarity by a margin of only 0.003-0.064. Qwen3-Embedding-4B, the strongest probe in the panel, reaches 91.3% probe accuracy but only 64.3 retrieval MAP. Because retrieval, clustering, deduplication, and retrieval-augmented generation operate on the metric rather than on a trained probe, these systems are effectively blind to information t

Collaborative In-Context Learning for Indoor Radio Map Estimation

arXiv:2610.31702v1 Announce Type: new Abstract: Indoor radio map estimation (RME) is essential for wireless applications such as coverage planning, localization, and network deployment. However, most existing methods follow a supervised learning paradigm that requires costly labeled data and task-specific model training. Moreover, indoor environments involve privacy-sensitive information, such as building layouts and structures, that often cannot be centralized. To address these challenges, we make the first attempt to formulate indoor RME as a collaborative in-context learning (ICL) problem. We propose Poly-Ripple, a framework in which multiple clients perform ICL using large autoregressive vision models (LAVMs) without model updates. Each client is prompted by its private demonstrations

Context or Knowledge: How LLMs Attribute Their Own Answers

arXiv:2610.32109v1 Announce Type: new Abstract: Language models can answer a question from a supplied document or from their parametric knowledge. When later asked where the answer came from, what determines their source report? We study this by intervening on the model's key-value (KV) cache after the answer is generated and before the source question is asked. On a controlled factual task with four models (7B-32B), changing or removing the document's cache never changes the report. In contrast, transplanting the question-and-answer (Q+A) cache between runs with identical question and answer tokens transfers it: every eligible report follows the donor, while the recipient's document is unchanged. On the two smaller models, steering Q+A values along directions fitted before any source ques