CoCoA: Collaborative Chain-of-Agents for Parametric-Retrieved Knowledge Synergy

CoCoA Framework:The top part is CoCoA-zero, a multi-agent collaboration framework. It integrates internal and external knowledge in a collaborative manner by first performing knowledge induction and then making decisions. The bottom part is the training strategy, which is based on CoCoA-zero and combines the trajectories of different agents into long chains to train and enhance the integration ability of the LLM.

Details will be completed soon ...

🛠 Installation

The main dependencies are torch 2.5.1, vllm 0.7.3, DeepSpeed, trl, peft, faiss/faiss-gpu.

conda create -n CoCoA python=3.9.18
conda activate CoCoA
pip install -r requirements.txt

💡 Preparation

Download Corpus & Index

Retrieval is performed on the set of Wikipeda passages used in DPR. Download passages:

wget https://dl.fbaipublicfiles.com/dpr/wikipedia_split/psgs_w100.tsv.gz

Download passage embeddings pre-computed with Contriever or Contriever-msmarco:

wget https://dl.fbaipublicfiles.com/contriever/embeddings/contriever/wikipedia_embeddings.tar
wget https://dl.fbaipublicfiles.com/contriever/embeddings/contriever-msmarco/wikipedia_embeddings.tar

Retrieve top-k passages:

cd ./retrieval
python retrieval_engine.py # Remember to configure your parameters

🎯 Train LLM

Training

cd scripts
bash xxx.sh # You can view the scripts provided in the scripts directory

📈 Run Evaluation

Download Evaluation Data:

HotpotQA, 2WikiMultiHopQA, WebQuestions, TriviaQA

Details will be completed soon

Citation

@article{jiang2025collaborative,
  title={Collaborative Chain-of-Agents for Parametric-Retrieved Knowledge Synergy},
  author={Jiang, Yi and Zhao, Sendong and Li, Jianbo and Wang, Haochun and Zhang, Lizhe and Liu, Yan and Qin, Bing},
  journal={arXiv preprint arXiv:2508.01696},
  year={2025}
}

Thanks for your interest in our work!