The Archive

MirrorCode benchmark evaluates AI coding agents on full program reimplementation tasks, extending beyond short coding exercises.

Introduces AgentCanvas, a framework automating embodied agent architecture search via typed-graph runtime, extending prior text-domain AAS to perceptual agents.

Jon Udell argues for 'agent-assisted' framing in software development: agents join human workflows rather than replacing human authority in autonomous loops.

AI agents have changed a lot in the last two years. The first could only answer one question at a time. Then came multi-turn chat, where the model could keep... AI agents have changed a lot in the last two years. The first could only answer one question at a time. Then came multi-turn chat, where the model could keep some context across a session. Today, we have long-horizon agents. Systems that plan many steps, split work between sub-agents, keep context across a long task, and run tools in a safe sandbox. The NVIDIA AI-Q Blueprint is an open source… Source

Fictional incident report imagining competing AI code review agents entering costly disagreement loops, illustrating multi-agent coordination and cost/security risks.

Agent-Native Immune System (ANIS): runtime defense framework for autonomous agents against memory poisoning, tool-chain manipulation, multi-agent attacks.

Study shows autonomous AI coding agents pass individual benchmarks but accumulate integration friction in shared codebases, revealing ecosystem-level risks.

CPAgents uses agentic iteration to auto-generate cardiac phenotypes for disease association studies via composite feature discovery.

Tandem RL training pairs weak and strong agents to maintain verifiable reasoning quality while improving compatibility and readability.

ToolPrivacyBench evaluates purpose-bound privacy in multi-tool LLM agent trajectories, filling gap in agent evaluation.

Cohere publishes practical guide on building AI agents for enterprise automation, covering reliability, security, and deployment patterns.

Agent testing startup Patronus AI, founded by former Meta AI researchers, is experienced nearly insatiable demand, its investor says.

Notion is "going all in on using agents to run your inbox."

PEEU method trains open-source MLLMs for GUI task planning via autonomous environment exploration and hindsight experience replay.

Study identifies co-failure ceiling limiting accuracy gains in multi-model LLM routing/voting systems across 67 frontier models; proposes beta metric beyond pairwise correlation.

Framework for persistent embodied agents combining cyber-physical action spaces with autonomous failure recovery in unstructured environments.

General Intuition has raised $320 million to scale AI trained on millions of hours of gameplay, betting action data can help AI develop something closer to human intuition.

Joint learning framework combines natural-language experiential rules and parameter updates to improve LLM agent policies in sparse-reward multi-step environments.

A new OpenAI research paper shows how AI agents are transforming work, enabling longer, more complex tasks and expanding productivity across roles.

AI agents automate vLLM fork maintenance, cutting upstream sync time from weeks to days by auto-rebasing, testing, and fixing conflicts.

Process reward models enable fine-grained, step-level evaluation of LLMs, yet building them for agentic settings remains prohibitively difficult: long-horizon interactions, irreversible actions, and stochastic environment feedback make both human annotation and Monte Carlo estimation infeasible at scale. In this work, we show that reinforcement learning (RL) post-training already provides the ingredients for effective step-level scoring, eliminating the need for dedicated reward model training altogether. Concretely, we derive an implicit advantage under a general stochastic Markov decision p...

AI agents are granted access to tools, APIs, and other infrastructure, making them active principals in those systems. The dominant approach places controls inside the agent's own runtime: system prompts, output filters, and guardrail libraries. Any control in the agent's address space is reachable by inputs that influence it; this generalizes to any AI system with sufficient reach into its own runtime, a class we term escapable AI systems. We identify four properties that an authorization mechanism must satisfy for architectural control rather than for cooperative requests: process separatio...

As autonomous AI agents increasingly transact across organizational boundaries, a fundamental trust challenge emerges: how can an agent assess whether an unknown counterpart is trustworthy? The ERC-8004 protocol addresses this challenge with the first permissionless trust layer for AI agent economies, built around three on-chain registries for Identity, Reputation, and Validation. Despite its rapid adoption, the protocol has not been studied empirically, leaving it unclear whether the information it records provides a trustworthy basis for decision-making. To address this gap, we present the ...

We present HiReLC, a hierarchical ensemble-reinforcement learning framework for automated joint quantization and structured pruning of deep neural networks. The framework decomposes the compression search across two levels of abstraction: low-level agents (LLAs) operate independently per block, selecting per-kernel configurations over a multi-discrete action space spanning bitwidth, pruning keep-ratio, quantization type, and granularity, while high-level agents (HLAs) coordinate global budget allocation via ensemble voting guided by Fisher Information-based sensitivity estimates. To mitigate ...

Figma has unveiled some new design and coding product updates at its annual Config conference that aim to help creatives "push their ideas further" and automate tedious tasks with AI. Part of this is a reimagined canvas that's now optimized for full-stack development, according to Figma, bringing teams, AI agents, tools, and materials "together in one place." Notable callouts include coding layers that let you tweak the code of your projects without leaving the Figma Design canvas, and AI-generated motion graphics effects, which allow you to make animations and transition effects by simply de...

We introduce Autodata, a general method that enables AI agents to act as data scientists who build high quality training and evaluation data. We show how to train (meta-optimize) such a data scientist agent, so that it learns to create even stronger data. We describe the overall formulation, and a specific practical implementation, Agentic Self-Instruct. We conduct experiments on computer science research tasks, legal reasoning tasks and reasoning with mathematical objects, where we obtain improved results compared to classical synthetic dataset creation methods. Further, meta-optimizing the ...

Multi-agent goal recognition asks an observer to jointly infer which agents act together and what each team is trying to achieve, so the hypothesis space grows combinatorially with the number of team partitions and goals per team. Real applications such as drone surveillance and collaborative robotics expose only the agents' trajectory, which forces the observer to rank team-goal hypotheses from behavior alone. Multi-Agent Goal Recognition with Branch-and-Bound (MAGR-BB) addresses this setting with a shared team- and goal-conditioned policy used as the scoring model inside a factorized branch...

Small ($\sim$2B) GUI-grounding agents are attractive for on-device deployment, accessibility tooling, and low-cost iteration, but at this scale they face two open recipe questions: how to obtain bounding-box training data without expensive human annotation, and how to combine supervised fine-tuning with reinforcement learning. We address both, with the explicit goal of pushing small-model performance rather than scaling up. WinDOM is a $54{,}425$-record grounding corpus harvested by driving an open-source Windows 11 web reimplementation under headless Playwright, with bounding boxes read dire...

Group-based reinforcement learning effectively post-trains LLM agents for long-horizon, sparse-reward tasks by deriving step-level credit from trajectory outcomes. However, this ties a step's credit to its rollout's final outcome: semantically near-identical intermediate steps receive opposite credit depending on whether their trajectory eventually succeeded or failed. Such semantic credit inconsistency sends conflicting gradients to similar actions and wastes the partially-correct progress inside failed rollouts. Motivated by this, we propose Semantic Consistency Policy Optimization (SCPO), ...

To better assist users with completing challenging tasks, AI agents mediate communications, access data, and interact with different APIs. Many employers (and even nation-states) already provide their users with this technology. However, widespread adoption of AI agents creates a new risk to abuse access to user data for another goal: surveilling users. These users might not even have the ability or permission to control the actions and data accesses of the surveilling agents. We introduce and formalize the problem of agentic surveillance: the ability of an AI agent to analyze available infor...