MikeTrendsTrends right now
AI: reposAPI

search

LLM agents

Trends

  1. 1
    New paper proposes hybrid reasoning agents for LLMs●Hybrid Reasoning Agents: LLM agents that think in latent space, act through textYhnScienceBiology52021 h ago

    A new arXiv paper, 'Hybrid Reasoning Agents', describes LLM agents that reason in latent space while acting through text. The work, ranking highly on Hacker News, argues that letting models think internally before emitting text could make agent behavior more efficient and capable, and readers are debating the implications for agent design.

  2. 2
    Open-source multi-agent AI trading framework gains attention●TauricResearch/TradingAgents⬢github1.4K15 min ago

    TauricResearch has released TradingAgents, a Python framework that uses multiple large language model agents to simulate financial trading decisions. The project, described as a multi-agent LLM financial trading framework, is drawing attention from developers and finance enthusiasts interested in applying AI agents to markets and automated trading research.

  3. 3
    New repository turns Karpathy's coding advice into Claude rules●multica-ai/andrej-karpathy-skills⬢github70215 min ago

    A GitHub repository called andrej-karpathy-skills offers a single CLAUDE.md file intended to improve Claude Code's behavior, built from Andrej Karpathy's published observations about common pitfalls in LLM-assisted coding. The project translates his critiques into practical instructions developers can drop into their coding-agent setup, and it has started gaining traction among developers experimenting with AI coding tools.

  4. 4
    Tsinghua researchers open-source AKV cache system for LLM agents●📰 Tsinghua AI Open-Sources AKV, a Cache Engineering Server for LLM Agents on SGLang Tsinghua researchers released AKV, aMmastodonTechnologyAI09 h ago

    Tsinghua researchers have released AKV, an open-source serving system built on SGLang that applies cache engineering to LLM agents. Instead of summarizing conversation history, AKV drops and repositions KV cache entries, reportedly cutting average serving costs by 27.7% across tested workloads.

  5. 5
    LLM Agents Now Play Pokémon Showdown Battles●LLM agents play Pokemon showdownYhn264 d ago

    A research project by Oliver Grabner demonstrates large language model agents competing in Pokémon Showdown battles, the online competitive battling simulator. The work explores how general-purpose language models handle turn-based strategy, incomplete information and move selection against human-style opponents, adding to a growing wave of experiments testing AI agents in game environments.

  6. 6
    New benchmark tests how AI agents handle mid-task disruptions●Benchmarking the responsiveness of LLM-based agentic systems under controlled mid-task perturbations✉newsTechnologyAI5 d ago

    Researchers have published a benchmark measuring how well LLM-based agentic systems respond to controlled perturbations introduced mid-task. The work, reported in Nature, evaluates whether autonomous AI agents can adapt when conditions change partway through an assignment, a key reliability concern as such systems move into real-world use. It is being discussed among AI researchers focused on agent robustness and evaluation standards.

Repos