LIVE PULSE
3.9 Anthropic CEO Amodei calls for slower AI development and shared safety rules — 11 src2.1 Agility Robotics unveils Digit 5 humanoid for warehouses and factories — 2 src2.0 Apple ships rebuilt Siri with Google Gemini, but not in the EU — 2 src1.7 Siri AI in macOS 27 Golden Gate: FAQ, Germany availability, privacy questions — 2 src1.4 Sam Altman says OpenAI will not go public in 2026, citing AI safety concerns — 5 src1.1 OpenAI contractors review real ChatGPT conversations to rate responses, report says — 2 src1.1 Anthropic data retention policy prompts firms to limit Claude use for sensitive work — 1 src1.1 Study examines issue bias in LLMs used as writing assistants before Swedish 2026 election — 1 src1.1 Study Audits Misalignment in Multi-Modal World Models — 1 src1.1 Retrieval-Grounded Reasoning Approach Proposed for Universal Multimodal Embeddings — 1 src3.9 Anthropic CEO Amodei calls for slower AI development and shared safety rules — 11 src2.1 Agility Robotics unveils Digit 5 humanoid for warehouses and factories — 2 src2.0 Apple ships rebuilt Siri with Google Gemini, but not in the EU — 2 src1.7 Siri AI in macOS 27 Golden Gate: FAQ, Germany availability, privacy questions — 2 src1.4 Sam Altman says OpenAI will not go public in 2026, citing AI safety concerns — 5 src1.1 OpenAI contractors review real ChatGPT conversations to rate responses, report says — 2 src1.1 Anthropic data retention policy prompts firms to limit Claude use for sensitive work — 1 src1.1 Study examines issue bias in LLMs used as writing assistants before Swedish 2026 election — 1 src1.1 Study Audits Misalignment in Multi-Modal World Models — 1 src1.1 Retrieval-Grounded Reasoning Approach Proposed for Universal Multimodal Embeddings — 1 src
HEATPULSEAI MAGAZINES
FLIP · FOLLOW · SAVE
Log inSign up

long-horizon tasks

topic6 events
papersTODAY 04:00 UTC

WISE: Long-Horizon Minecraft Agent with Why-Which Reasoning

A revised arXiv paper introduces WISE, an approach for long-horizon tasks in Minecraft that combines LLM-driven hierarchical control with a "why-which" reasoning scheme. The authors note that in existing LLM-augmented hierarchical agents, the low-level controllers frequently limit overall performance, which their method aims to address. The work targets general-purpose embodied agents operating over extended task horizons.

agents/minecraft/embodied-ai/research
papersTODAY 04:00 UTC

DynSTEER: Dynamic Stage-wise Evaluation and Review for LLM Agents

A new arXiv paper introduces DynSTEER, a framework for evaluating large language model agents that operate over long-horizon tasks. It targets gaps in existing evaluation methods, which typically judge only final outcomes and struggle to pinpoint where errors occur. The approach combines stage-wise trajectory assessment with review during execution rather than after the fact.

agents/evaluation/llm-agents/research
papersSEP 12 04:00 UTC

T1: 122B Mixture-of-Experts Model Trained with RL for Terminal Agent Tasks

Researchers released T1, a 122-billion-parameter Mixture-of-Experts model trained via reinforcement learning to act as an agent in terminal environments. The work targets long-horizon workloads such as software development and scientific research, where sustained multi-step command-line use matters. It is presented as part of a broader shift in agent design away from short, single-turn interactions.

agents/reinforcement-learning/mixture-of-experts/long-horizon
papersSEP 12 04:00 UTC

Mr.LHDR benchmark evaluates multimodal long-horizon deep research agents

A new arXiv paper introduces Mr.LHDR, a benchmark designed to test deep research agents on extended, multi-step tasks. The authors note that current benchmarks mostly measure shorter exploratory work and seldom assess whether agents can keep going over long horizons. It focuses on web search, tool use and combining evidence from multiple modalities.

benchmarks/agents/multimodal/research
papersSEP 10 04:00 UTC

Reinforcement Learning with Temporal-Logic-Based Causal Diagrams

A revised arXiv paper (v2) studies reinforcement learning problems in which agents must achieve goals that unfold over long time horizons, a setting commonly handled by encoding tasks as deterministic finite automata. The authors propose representing these tasks with causal diagrams built from temporal logic, aiming to give agents a structured way to reason about extended objectives. The update is a replacement version of the original June 2023 preprint.

reinforcement-learning/temporal-logic/causality/planning
papersSEP 10 04:00 UTC

Paper Compares Subagents and Agent Skills for Long-Horizon Agentic Tasks

A new arXiv paper investigates how language model agents can draw on libraries of reusable knowledge when tackling long-horizon tasks. It contrasts two approaches—subagents and agent skills, where skills are packaged as multi-file bundles—and examines which executes such knowledge more effectively. The study was announced in the cs.AI category and cross-listed in cs.CL and cs.LG.

agents/llm/skills/long-horizon