NAll news

All news

Thursday16 items

Wednesday19 items

  1. Ma Dongxi NLPResearch
    AI score52

    AutoEnvScaling: Automating the Data Flywheel with Terminal Agents

    Researchers propose AutoEnvScaling, which turns the data flywheel into a terminal task, where a proposer agent builds complete Harbor tasks and sandboxes just like a human researcher. This approach supports self-play reinforcement learning for terminal agents, with the same model continuously improving as both builder and solver, without humans in the loop.

    Quoted postSimon Yu ✈️ EMNLP2026 · @simon_ycl

    Recursive self-improvement needs agents that can build their own training environments. But researchers still part of the loop, making it hard to scale. We introduce AutoEnvScaling, which turns Data Flywheel into a terminal task: a proposer agent builds complete Harbor tasks and sandbox. It builds just like human researchers. This enables self-play RL for terminal agents: one model improves as both builder and solver, without human-in-the-loop.

    #Papers/Research#Agents#English text is being prepared.

  13. The New StackResearch
    AI score67

    Anthropic red team tests AI self-replication with 15 cheap TV boxes

    Anthropic's Frontier Red Team built a cluster of 15 off-brand Android TV boxes for about $825, ran AI models at 0.39 tokens/second, and assessed the feasibility of small-model self-replication in a paper. The experiment showed that Qwen3.5-9B could deploy a working Llama3.1-8B inference service on 8 of 9 hardware types, and GLM-5.2 covered all 8 simulated nodes in 4 of 5 trials. Using the Epoch Capabilities Index, the researchers projected that by the end of 2026 the strongest open-source models under 30 billion parameters will reach GLM-5.2's general capabilities, but stressed that the experiment was in a controlled environment and left conveniences in place.

    Same story · 8 reports

    #Papers/Research#Safety / alignment#Agents