EvoDS: Self-Evolving Autonomous Data Science Agent with Skill Learning and Context Management

🎉 Accepted by the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 2026)

---

📖 Overview

EvoDS is a self-evolving autonomous data science agent designed to address two fundamental challenges in LLM-based data science systems:

1. Reusable Skill Learning — enabling agents to synthesize, validate, and internalize reusable tool-usage skills from experience.
2. Adaptive Context Management — dynamically compressing long-horizon interaction history to maintain stable reasoning under limited context budgets.

EvoDS formulates autonomous data science as a sequential decision-making process over an evolving action space under bounded context constraints. The framework integrates three key components:

🔧 Autonomous Skill Acquisition (ASA)

ASA treats tools as learnable capabilities. The agent can autonomously synthesize, validate, cache, and reuse executable skills during task solving.

🧠 Adaptive Context Compression (ACC)

ACC dynamically determines when and how to compress interaction history during long-horizon multi-step reasoning, improving context efficiency and reasoning stability.

🔁 Agentic Reinforcement Learning

EvoDS jointly optimizes:

• task completion quality,
• skill acquisition behavior,
• and context regulation policies

through joint reinforcement learning over multiple agent roles.

---

📦 Checkpoints

We provide pretrained checkpoints for EvoDS.

EvoDS Checkpoint

Download the checkpoint from:

• Hugging Face - EvoDS

Place the downloaded files under:

checkpoints/EvoDS

---

🚀 Quick Start

🔑 Step 1: Configure API Key

Edit:

config/config.yaml

Replace:

<your_openai_key>

with your OpenAI API key.

---

📚 Step 2: Prepare Benchmarks

Due to storage constraints, benchmark datasets are not included in this repository.

Please manually download the following benchmarks before evaluation:

• DABench
• DA-Code
• ScienceAgentBench
• MLE-Dojo

After downloading, place the datasets in the corresponding benchmark directories specified in the configuration files.

---

▶️ Step 3: Launch EvoDS

Run:

bash run.sh

This script will:

1. Deploy EvoDS using vLLM
2. Sequentially evaluate the agent on:

• DABench
• DA-Code
• ScienceAgentBench
• MLE-Dojo

---

🖥 Multi-GPU Configuration

By default, run.sh uses a single GPU:

export CUDA_VISIBLE_DEVICES=0

To use multiple GPUs:

1. Modify:

export CUDA_VISIBLE_DEVICES=0,1,2,...

2. Update the GPU count in:

start_llm "checkpoints/EvoDS" "EvoDS" 1234 1 "nohup.log"

Replace 1 with the number of GPUs you intend to use.

---

🏋️ Training

We use VERL for both supervised fine-tuning (SFT) and reinforcement learning (RL).

---

🔧 Before Training

Edit:

verl/verl/workers/reward_manager/datascience.py

Replace:

your_api_key

with your OpenAI API key.

---

📂 Enter the Training Directory

cd verl

Download Qwen3-8B and place it under:

Qwen/Qwen3-8B

---

🧠 Supervised Fine-Tuning (SFT)

Run:

bash examples/sft/multiturn/run_evods_qwen3_8b_multi_turn.sh 4 checkpoints/evods_sft_8b

Arguments:

• 4 — number of GPUs
• checkpoints/evods_sft_8b — output checkpoint directory

The trained checkpoint will be saved to:

checkpoints/evods_sft_8b

---

🔁 Reinforcement Learning (RL)

Run:

bash examples/sglang_multiturn/run_evods_qwen3_8b.sh

This stage performs multi-agent reinforcement learning to jointly optimize:

• task completion quality
• tool scheduling efficiency
• context length control
• skill acquisition behavior

---