NTU Part II, pages 46–50. Seven worked examples connect SEAL self-edit rewards, two timescales of learning, learned optimizers and MAML. Follow the inner and outer updates and check each reported result against its evaluation unit.
Topic: #ai-agents
A collection of 14 posts about ai-agents.
NTU Part II, pages 41–45. Seven worked examples connect gradient descent, PostTrainBench, autoresearch and program evolution. Calculate what changes when an agent edits a training algorithm.
NTU Part II, pages 36–40. Follow an edit across generations, compare three parent-selection rules in Hyperagents, and calculate how Learning to Self-Evolve rewards an editor. Seven worked examples separate task performance from editing skill.
Pages 31–35 of NTU’s second lecture separate task success from skill retention. Six worked examples trace harness edits, capability-preserving evolution, and the editor in the Darwin Gödel Machine.
Pages 21–25 connect ExpRAG, BetterTogether and E-SPL. Papers and Omar Khattab’s lecture explain experience retrieval, alternating optimization and reevaluation when goals change.
Pages 16–20 of the NTU lecture explore DGM’s search archive, declarative DSPy programs, harness and model parameters, experience retrieval and fine-tuning.
Pages 11–15 examine OPRO, GEPA, ALMA and the Darwin Gödel Machine. Slides and author commentary illuminate prompt search, memory design and code self-modification.
Pages 6–10 connect prompt, tool and workflow changes to SWE-agent, DSPy, AFlow and τ-bench. The discussion covers candidate selection, memory, cost and repeatability.
ReAct, Generative Agents, Reflexion and SWE-agent show how weights, harness design and runtime state affect behavior and what makes an evaluation fair.
A Science study asks models to choose questions and tests in sequence. Its reported 83.2% accuracy raises specific questions for clinical validation.
LEASH uses uncertainty in hidden states to escalate selected computer-use tasks. The preprint compares cost and success on OSWorld and identifies gaps in validation.
Google Research combined lab-state tracking, simulation and execution logs to identify liquid-handling errors and correct experiments in an early demonstration.
The editorial workflow covers checking sources, equations and figures, parallel review, version control, approval and verification after publication. The article identifies the steps already tested and the remaining cloud and Blogger integration work.
Read OpenAI’s research automation data through its original diagrams and equations. The analysis covers agent roles, runtime, productivity, limits to parallel work, human intervention, and the denominator behind success rates.