Llm Tuning Patterns

by parcadeid07ff4b06b62No license3.9K starsListed Oct 8, 2026Updated Oct 8, 2026Repository updated 8 months ago

LLM Tuning Patterns

Instructions onlyAI & Agents

AI-generated overview

Reference patterns for configuring LLM parameters such as max_tokens, temperature and top_p by task type.

What it does: This skill provides evidence-based guidance for setting LLM generation parameters according to the task at hand. It gives parameter tables for theorem proving and formal reasoning, code generation, and creative or exploration tasks, each with stated rationales. It also lists anti-patterns and a proof-plan prompt template for chain-of-thought before tactics.
When to use it: Use it when choosing max_tokens, temperature or top_p values for an LLM task, especially theorem proving, code generation or creative work. It is also useful for reviewing common misconfigurations such as too few tokens or too low temperature for proofs.
Requirements: No tools, packages or credentials are required; it is instructions only and ships no scripts.

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:

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.

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.

Parallel Sampling

For hard proofs, use parallel sampling:

Generate N=8-32 candidate proof attempts
Use best-of-N selection
Each sample at temperature 0.6-0.8

Code Generation

Creative / Exploration Tasks

Anti-Patterns

Too low tokens for proofs: 512 tokens truncates chain-of-thought
Too low temperature for proofs: 0.2 misses creative tactic paths
No proof plan: Jumping to tactics without planning reduces success rate

Source Sessions

This session: APOLLO parity - increased max_tokens 512->4096, temp 0.2->0.6
This session: Added proof plan prompt for chain-of-thought before tactics

Source and attribution

Source:parcadei/continuous-claude-v3in.claude/skills/llm-tuning-patternsat commitd07ff4b

License: No license

Content belongs to its original authors. SourceWeft indexes it from a public repository.

Report or request removal

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

Parameter	Value	Rationale
max_tokens	2048	Sufficient for most functions
temperature	0.2-0.4	Prefer deterministic output

Parameter	Value	Rationale
max_tokens	4096	Space for exploration
temperature	0.8-1.0	Maximum creativity

Llm Tuning Patterns

LLM Tuning Patterns

Pattern

Theorem Proving / Formal Reasoning

Proof Plan Prompt

Parallel Sampling

Code Generation

Creative / Exploration Tasks

Anti-Patterns

Source Sessions

Source and attribution

More from parcadei/continuous-claude-v3

Tldr Stats

Wiring

Workflow Router

Validate Agent

Tour

Tldr Overview

More in AI & Agents

Ppt Template Creator

Skill Development

Plugin Structure

Command Development

Claude Md Improver

Find Skills