recursive-reasoner-planner-llama3.1-8b

Recursive Reasoner Planner

A fine-tuned version of meta-llama/Llama-3.1-8B-Instruct for planning in recursive reasoning systems.

Purpose

This model serves as a planner that analyzes math problems and decides whether to:

1. Decompose the problem into subproblems, or
2. Solve directly (atomic)

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("vkaarti/recursive-reasoner-planner-llama3.1-8b")
tokenizer = AutoTokenizer.from_pretrained("vkaarti/recursive-reasoner-planner-llama3.1-8b")

prompt = '''Analyze this math problem and decide how to solve it:

Problem: Janet's ducks lay 16 eggs per day. She eats three for breakfast every morning and bakes muffins for her friends every day with four. She sells the remainder at the farmers' market daily for $2 per fresh duck egg. How much in dollars does she make every day at the farmers' market?

Your task:
1. Decide if this problem should be decomposed into subproblems or solved directly
2. If decomposing: identify 2-3 distinct calculation steps as subproblems
3. Explain your approach briefly

Return ONLY valid JSON:
{
    "should_decompose": true/false,
    "subproblems": ["step 1", "step 2", ...] or [] if atomic,
    "plan": "Brief explanation of approach"
}'''

inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Output Format

The model outputs JSON with:

• should_decompose: Boolean indicating whether to break down the problem
• subproblems: List of subproblem descriptions (if decomposing)
• plan: Brief explanation of the approach