Tag: research-brief

Showing 1-2 of 2

Feb 06, 20266 min readFounder NotesIntermediate5 min build

State-level selective verification with learned heuristics for verification-cost-limited LLM reasoning

Examines a state-level selective verification pipeline—feasibility gating, learned scoring and ranking, and adaptive verifier allocation—that trims verifier calls by 44% on MATH.

LLM verification compute-allocation inference-optimization test-time

+2 more

MATH-benchmark research-brief

Model BreakdownsUnited Kingdom

Open

Feb 06, 20266 min readFounder NotesIntermediate5 min build

Task-Method-Knowledge Prompting Improves LLM Planning on PlanBench Blocksworld

TMK prompts (Task / Method / Knowledge) raised LLM planning accuracy on PlanBench Blocksworld from 31.5% to 97.3%. Practical steps and reproduction tips for builders.

LLM prompting TMK planning PlanBench

+3 more

Blocksworld chain-of-thought research-brief