Llm Tuning Patterns

A collection of evidence-based LLM parameter configurations drawn from formal reasoning research like APOLLO and Godel-Prover. The core insight is that theorem proving needs different settings than code generation: higher temperature (0.6 vs 0.2-0.4) and more tokens (4096 vs 2048) because proofs benefit from creative tactic exploration and chain-of-thought reasoning. The proof plan prompt pattern is smart, asking the model to sketch its approach before diving into tactics. If you're doing formal verification or mathematical reasoning with LLMs, these tuning patterns are worth adopting. The anti-patterns section is especially useful, calling out common mistakes like truncating proofs with low token limits.

skills.sh installs: 480
GitHub stars: 3.9K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add parcadei/continuous-claude-v3 --skill llm-tuning-patterns --agent claude-code

Installs into .claude/skills of the current project.

Files

LLM Tuning Patterns

Evidence-based patterns for configuring LLM parameters, based on APOLLO and Godel-Prover research.

Pattern

Different tasks require different LLM configurations. Use these evidence-based settings.

Theorem Proving / Formal Reasoning

Based on APOLLO parity analysis:

Parameter	Value	Rationale
max_tokens	4096	Proofs need space for chain-of-thought
temperature	0.6	Higher creativity for tactic exploration
top_p	0.95	Allow diverse proof paths

Proof Plan Prompt

Always request a proof plan before tactics:

Given the theorem to prove:
[theorem statement]

First, write a high-level proof plan explaining your approach.
Then, suggest Lean 4 tactics to implement each step.

The proof plan (chain-of-thought) significantly improves tactic quality.