Results for “agents”

…echoed by @corbtt, who notes that customers using RL to train agents on specific domains are "extremely happy." The broader challenge, as @jxmnop describes it, is the palpable tension among researchers to make post-training as "clean and…

…which make it much better/faster/cheaper for long running agents: We are so glad that DeepSeek is back publishing SOTA research. Our last highlight is their comments on post-training, where they largely seem to agree with Prof Jie Tang:…

…AI-powered side projects using GPT-4 and GPT-3.5, and interface agents powered by LLMs. In the Stable Diffusion space, users seek realistic SDXL models comparable to PonyXL, praise the Hi-diffusion extension for ComfyUI, and note the…

…orchestration to multi-agent products with offerings like Deep Agents/Open SWE and LangSmith Fleet. The discourse emphasizes the importance of clear base-model attribution, licensing compliance, and product differentiation through…

…a quiet day lets us report on a long trend of the major coding agents

…and potential disruption of the call center industry by AI agents.

…risk as capabilities rise (@METR_Evals). Reasoning/agents: GPT‑5 shows strong instruction following and tool use (e.g., TauBench gains, IFBench instruction‑following), but mixed changes on non‑SWE coding evals and OpenAI PR‑reproduction…

…to schedule reminders and summaries, and an open-source ambient agents email assistant. OpenAI rolled out Tasks for scheduling actions in ChatGPT for Plus, Pro, and Teams users. AI software engineering is rapidly advancing, predicted to…

…and “Gemini Enterprise” as a no-code front door to build agents and automate workflows across Workspace/M365/Salesforce and more (@Google, @JeffDean). Memory and eval-driven optimization in agent pipelines: Developers test memory layers…

…using efficient exploration was articulated, with a focus on agents that craft queries and a reward model that functions on the feedback received. The discussion encompassed the merits of double Thompson sampling and the employment of…

…example generation compared to SFT, allowing them to train agents from small OSS models that outperform frontier models on specific tasks. Critique and Improvement of AI Agent Benchmarks: A blog post shared by @ShayneRedford and work from…

…GEAR team discussed Project GR00T, aiming to develop AI agents capable of operating in both simulated and real-world environments, enhancing generalist abilities in robotics. OpenAI's Commitment to Safe AGI Development: Members highlighted …