LRPlan

This is the official repository for the EMNLP 2025 Findings paper:
LRPLAN: A Multi-Agent Collaboration of Large Language and Reasoning Models for Planning with Implicit & Explicit Constraints

LRPlan is a multi-agent architecture designed to solve complex real-world planning problems.

)

Setup

1. Create a conda environment and install dependencies:

conda create -n lrplan python=3.9
conda activate lrplan
pip install -r requirements.txt

2. (Only needed for TravelPlanner) Download the database and unzip it to the Code/TravelPlanner/feedback_script/travelplanner/ directory (creating a folder named 'database' inside Code/TravelPlanner/feedback_script/travelplanner/).

Running

TravelPlanner

1. Write API Keys in OAI_CONFIG_LIST
2. Navigate to the directory Code/TravelPlanner/
3. Generate reasoning traces:
   1. Write your DeepSeek API key at line 33 in the file get_reasoning_traces.py
   2. Navigate back to the main directory.

   python get_reasoning_traces.py

4. Run Inference file in directory Code/TravelPlanner/LRPlan:

python LRPlan.py 

TimeArena-Static

1. Similar to TravelPlanner in the Code/TimeArenaStatic/ directory.

Prompts

1. All prompts are found in the directory Prompts/.
2. cot_instructions.py contains basic CoT instructions used for baseline comparisons.
3. pattern_extractor_corrector.py contains the prompt for meta-agents used in LRPlan.
4. timearena.py and travelplanner contain task description, output format, and a few sample input-output pairs for their respective datasets.

Data

1. Directory Data/ contains all query samples for both datasets.