CoALM-70B

CoALM-70B: Conversational Agentic Language Model

Model Description

CoALM-70B is our middle scale Conversational Agentic Language Model, designed to integrate Task-Oriented Dialogue (TOD) capabilities with Language Agent (LA) functionalities at a larger scale than its predecessor CoALM-8B. By leveraging CoALM-IT, a multi-task dataset interleaving multi-turn ReAct reasoning with complex API usage, CoALM-70B achieves state-of-the-art performance across TOD and function-calling benchmarks.

CoALM-70B has been fine-tuned on a comprehensive multi-tasking covering dialogue state tracking, function calling, and multi-turn reasoning, surpassing even proprietary models like GPT-4o on major conversational evaluation benchmarks: MultiWOZ 2.4 (TOD), BFCL V3 (LA), and API-Bank (LA).

Model Sources

<!-- Provide the basic links for the model. -->

• 📝 Paper: https://arxiv.org/abs/2502.08820
• 🌐 Project Page: https://emrecanacikgoz.github.io/CoALM/
• 💻 Repository: https://github.com/oumi-ai/oumi/tree/main/configs/projects/coalm
• 💎 Dataset: https://huggingface.co/datasets/uiuc-convai/CoALM-IT

---

Model Details

• Model Name: CoALM-70B
• Developed by: Colloboration of UIUC Conversational AI LAB and Oumi
• License: cc-by-nc-4.0
• Architecture: Fine-tuned Llama 3.3 70B Instruct
• Parameter Count: 70B
• Training Data: CoALM-IT
• Training Type: Full Fine-tunning (FFT)
• Fine-tuning Framework: Oumi
• Training Hardware: 8 NVIDIA H100 GPUs
• Training Duration: ~24 hours
• Evaluation Benchmarks: MultiWOZ 2.4, BFCL V3, API-Bank
• Release Date: February 5, 2025

---

Capabilities and Features

🗣 Conversational Agentic Abilities

• Multi-turn Dialogue Mastery: Handles long conversations with accurate state tracking.
• Advanced Function Calling: Dynamically selects and executes API calls for task completion.
• Enhanced ReAct-based Reasoning: Integrates structured reasoning (User-Thought-Action-Observation-Thought-Response).
• Zero-Shot Generalization: Excels in unseen function-calling and TOD tasks.

🚀 Benchmark Performance

• MultiWOZ 2.4 (TOD): Strong performance in dialogue state tracking and task success.
• BFCL V3 (LA): Superior function-calling abilities compared to language agents.
• API-Bank (LA): High accuracy in API call generation and response synthesis.

---

Training Process

🔧 Fine-tuning Stages

1. TOD Fine-tuning: Optimized for dialogue state tracking (e.g., augmented SNIPS in instruction-tuned format).
2. Function Calling Fine-tuning: Trained to generate precise API calls from LA datasets.
3. ReAct-based Fine-tuning: Enhances multi-turn conversations with API integrations through structured reasoning.

🔍 Training Hyperparameters

• Base Model: Llama 3.3 70B Instruct
• LoRA Config: Rank = 16, Scaling Factor = 32
• Batch Size: 7
• Learning Rate: 4e-5
• Optimizer: AdamW (betas = 0.9, 0.999, epsilon = 1e-8)
• Precision: Mixed precision (bfloat16)
• Warm-up Steps: 24
• Gradient Accumulation Steps: 1

---

💡 CoALM-IT Dataset

<img src="table.png" alt="CALM-IT Dataset Statistics" width="800"/>

---

📊 Benchmark Performance

<img src="results.png" alt="CALM-IT Dataset Statistics" width="1000"/>

Usage

🏗 How to Load the Model using HuggingFace

from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("uiuc-convai/CoALM-70B")
model = AutoModelForCausalLM.from_pretrained("uiuc-convai/CoALM-70B")

🛠 Example Oumi Inference

pip install oumi

# See oumi_infer.yaml in this model's /oumi/ directory.
oumi infer -i -c ./oumi_infer.yaml

🛠 Example Oumi Fine-Tuning

pip install oumi

# See oumi_train.yaml in this model's /oumi/ directory.
oumi train -c ./oumi_train.yaml

---

• Scalability to CoALM-405B: Next iteration will extend capabilities for even larger-scale conversations.
• Continuous Open-Source Expansion: Ongoing release of datasets, model weights, and training artifacts to foster community research.

---

Acknowledgements

We'd like to thank the Oumi AI Team for collaborating on training the models using the Oumi platform on Together AI's cloud.

License

This model is licensed under Creative Commons NonCommercial (CC BY-NC 4.0).

---

Citation

If you use CoALM-70B in your research, please cite:

@misc{acikgoz2025singlemodelmastermultiturn,
      title={Can a Single Model Master Both Multi-turn Conversations and Tool Use? CoALM: A Unified Conversational Agentic Language Model}, 
      author={Emre Can Acikgoz and Jeremiah Greer and Akul Datta and Ze Yang and William Zeng and Oussama Elachqar and Emmanouil Koukoumidis and Dilek Hakkani-Tür and Gokhan Tur},
      year={2025},
      eprint={2502.08820},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2502.08820}, 
}

For more details, visit Project Repository or contact [email protected].