Live feed

redwoodresearch.org · 2026-10-25 · Ranked: agent match · community signal · fresh 0.99 · score 3.35 · Context

Red-teaming AI agent monitors with an LLM that fakes tool outputs

github.com · 2026-10-25 · Ranked: agent match · community signal · fresh 0.96 · score 2.97 · Context

A 32K-window proxy that beat full context on long agent histories

github.com · 2026-10-25 · Ranked: agent match · community signal · fresh 0.97 · score 2.91 · Context

Show HN: The fastest, most token-efficient iOS Simulator MCP for AI agents

arxiv.org · 2026-10-22 · Ranked: agent + eval match · research watch · fresh 0.66 · score 2.47

SCOPE: Compiling Natural-Language Policies into Neurosymbolic Predicate Programs

Policy enforcement for LLM agents has shifted from single-turn content moderation to compound, application-specific policies evaluated over multi-turn transcripts and tool-call trajectories. Fine-tuned guardrail model... Context & related coverage →

arxiv.org · 2026-10-22 · Ranked: agent + evaluation match · research watch · fresh 0.66 · score 2.13

Post-Audit Mirage: Identifiability Limits of Offline Agent Verification

Offline tests can fail to reveal whether an agent update will remain safe after deployment. We operationalize this post-audit mirage using matched safe and harmful lifecycles that expose identical verifier-visible evi... Context & related coverage →

arxiv.org · 2026-10-22 · Ranked: agent + eval match · research watch · fresh 0.67 · score 1.99

The Value of Memory in Language Agents

Memory enables language agents to carry useful distinctions from past experience into future decisions. Its value depends on which distinctions survive, whether a bounded agent can use them, and what can be recovered... Context & related coverage →

arxiv.org · 2026-10-22 · Ranked: evaluation match · research watch · fresh 0.68 · score 1.75

Simulated Control Evals

Control evaluations measure whether safeguards can prevent an AI model from causing undesired outcomes even when it is intentionally trying to do so. These evaluations typically use purpose-built control settings and... Context & related coverage →

arxiv.org · 2026-10-22 · Ranked: evaluation match · research watch · fresh 0.68 · score 1.74

Cost-Aware Abstention for LLM-Based IT Ticket Classification

Wrong automatic routing of an IT support ticket can cost more than human triage. We evaluate cost-aware abstention as an offline, cost-conditional readiness assessment: select a routing policy on calibration data unde... Context & related coverage →

modal.com · 2026-10-22 · Ranked: practitioner analysis · fresh 0.53 · score 1.26

Introducing our Training API

Spindle is an open-source API for training and fine-tuning open models — deploy a server with a single command and run any Tinker-compatible training code against it. Context & related coverage →

cloud.google.com · 2026-10-22 · Ranked: agent match · cloud platform update · fresh 0.46 · score 0.66

Expanding Claude availability and deployment choice on Google Cloud

Every organization has unique requirements for building and scaling generative AI. At Google Cloud, we are committed to giving developers model choice and architectural flexibility to run workloads however they need.... Context & related coverage →