sadd:tree-of-thoughts

Tree of thoughts reasoning systematically explores multiple solution directions, prunes weak candidates using generated rubrics, and synthesizes the best final approach through multi-perspective judging and evidence-based evaluation.

Multi-agent evaluation uses several independent judges to assess proposals against tailored rubrics, reducing bias and increasing coverage of failure modes to consistently rank and select the strongest solution candidates.

A meta-judge workflow is needed for complex tasks requiring systematic search over a solution space, where generating evaluation rubrics, coordinating parallel agents, and adaptive strategy selection produce superior outcomes.

Evidence-based synthesis combines superior sections from competing solutions while addressing weaknesses identified during multi-agent evaluation, selecting an adaptive outcome like select-and-polish, redesign, or full synthesis.

No, this workflow operates independently without external dependencies, using internal multi-agent coordination and meta-judge generated YAML to orchestrate the exploration, pruning, expansion, and evaluation processes.

Yes, adaptive strategy selection chooses between select-and-polish, redesign, or full synthesis based on judge consensus and scoring patterns, allowing weak candidates to be redesigned or combined into a stronger output.