# AI-Scientist-v2 | Sakana AI | Knowledge Base | Instagit

The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search

GitHub Stars: 3k

Repository: https://github.com/SakanaAI/AI-Scientist-v2

---

## Articles

### [Benefits of Using Agentic Tree Search in AI Scientist v2](/SakanaAI/AI-Scientist-v2/benefits-agentic-tree-search-ai-scientist-v2)

Discover agentic tree search benefits in AI Scientist v2. Explore a parallel, self-correcting system for efficient AI research and automated pruning.

- Tags: deep-dive
- Published: 2026-03-28

### [How AI Scientist v2 Uses Semantic Scholar for Literature Search and Automated Citations](/SakanaAI/AI-Scientist-v2/how-ai-scientist-v2-uses-semantic-scholar)

Discover how AI Scientist v2 leverages Semantic Scholar for literature search and automated citations. Explore autonomous discovery and effortless manuscript generation.

- Tags: how-to-guide
- Published: 2026-03-28

### [AI Scientist v1 vs v2: How the Template-Driven System Evolved into an Agentic Research Engine](/SakanaAI/AI-Scientist-v2/difference-ai-scientist-v1-vs-v2)

Explore the evolution from AI Scientist v1's rigid templates to v2's agentic research engine. Discover how the Experiment Manager dynamically drives scientific discovery.

- Tags: deep-dive
- Published: 2026-03-28

### [How to Perform Ideation with AI Scientist v2: Automated Research Generation Guide](/SakanaAI/AI-Scientist-v2/how-to-perform-ideation-ai-scientist-v2)

Learn how to perform ideation with AI Scientist v2 for automated research generation Create proposals using LLMs literature search and multi-round reflection Get structured JSON outputs from this powerful tool

- Tags: how-to-guide
- Published: 2026-03-28

### [Default Timeout for Experiments in AI Scientist v2: Complete Configuration Guide](/SakanaAI/AI-Scientist-v2/default-timeout-experiments-ai-scientist-v2)

Discover the default experiment timeout in AI Scientist v2. Learn how to configure this setting using bfts_config.yaml for optimal experiment management.

- Tags: how-to-guide
- Published: 2026-03-28

### [How AI-Scientist-v2 Ensures Experiment Workspace Isolation: A Complete Technical Guide](/SakanaAI/AI-Scientist-v2/how-system-ensures-isolation-experiment-workspaces)

AI-Scientist-v2 ensures experiment isolation with unique directories and process subfolders, preventing file system collisions for concurrent runs. Learn how.

- Tags: how-to-guide
- Published: 2026-03-28

### [Understanding the Purpose of ai_scientist/llm.py in AI-Scientist-v2](/SakanaAI/AI-Scientist-v2/purpose-of-ai_scientist_llm.py)

Discover the central role of ai_scientist/llm.py in AI-Scientist-v2. This module unifies LLM interactions, manages client creation, routing, retries, and output parsing for seamless AI development.

- Tags: deep-dive
- Published: 2026-03-28

### [How to Specify Different LLMs for Different Tasks in AI Scientist v2](/SakanaAI/AI-Scientist-v2/specify-different-llms-for-tasks-ai-scientist-v2)

Learn how to specify different LLMs for different tasks in AI Scientist v2. Easily assign LLMs to code generation, feedback, write-ups & more using YAML config and command-line flags.

- Tags: how-to-guide
- Published: 2026-03-28

### [How to Run the Full AI Scientist v2 Pipeline: A Complete End-to-End Guide](/SakanaAI/AI-Scientist-v2/how-to-run-full-pipeline-ai-scientist-v2)

Run the full AI Scientist v2 pipeline end-to-end. Execute ideation with perform_ideation_temp_free.py and launch the autonomous workflow via launch_scientist_bfts.py for research, experimentation, and review.

- Tags: how-to-guide
- Published: 2026-03-28

### [What LLM Models Are Supported by AI Scientist v2: Complete Model Catalog](/SakanaAI/AI-Scientist-v2/llm-models-supported-ai-scientist-v2)

Discover which LLM models AI Scientist v2 supports. Explore the full catalog featuring models from Anthropic, OpenAI, Meta, Google, and more. Access the complete list now.

- Tags: api-reference
- Published: 2026-03-28

### [How AI Scientist v2 Gathers Citations for Research Papers: A Deep Dive into Semantic Scholar Integration](/SakanaAI/AI-Scientist-v2/how-ai-scientist-v2-gathers-citations)

Discover how AI Scientist v2 gathers research paper citations using Semantic Scholar integration. Learn about metadata retrieval, citation sorting, and reference formatting for LLM ideation.

- Tags: deep-dive
- Published: 2026-03-28

### [Difference Between perform_writeup.py and perform_icbinb_writeup.py in AI-Scientist-v2](/SakanaAI/AI-Scientist-v2/difference-between-perform_writeup-and-perform_icbinb_writeup)

Understand the differences between perform_writeup.py and perform_icbinb_writeup.py in AI-Scientist-v2. Generate LaTeX papers for ML conferences and ICBINB workshops with specific optimizations.

- Tags: internals
- Published: 2026-03-28

### [How to Generate Research Papers with AI Scientist v2: End-to-End Workflow Guide](/SakanaAI/AI-Scientist-v2/how-to-generate-research-papers-ai-scientist-v2)

Learn how to generate research papers with AI Scientist v2. This guide details the end-to-end workflow for this autonomous research system, from ideation to manuscript creation.

- Tags: how-to-guide
- Published: 2026-03-28

### [How Ablation Studies Are Performed in Stage 4 of AI Scientist v2](/SakanaAI/AI-Scientist-v2/how-ablation-studies-performed-stage-4-ai-scientist-v2)

Discover how SakanaAI's AI Scientist v2 performs ablation studies in Stage 4. Learn how the ParallelAgent class isolates component impact and refines AI models.

- Tags: how-to-guide
- Published: 2026-03-28

### [What Datasets Are Used in Stage 3 of AI Scientist v2: Multi-Dataset Evaluation Explained](/SakanaAI/AI-Scientist-v2/datasets-used-in-stage-3-ai-scientist-v2)

Discover the datasets used in Stage 3 of AI Scientist v2. Learn how standard benchmarks and synthetic variants are evaluated for robustness and visualized.

- Tags: deep-dive
- Published: 2026-03-28

### [AI-Scientist-v2 Hyperparameter Tuning in Stage 2: Closed-Loop Baseline Optimization](/SakanaAI/AI-Scientist-v2/how-system-handles-hyperparameter-tuning-stage-2)

Discover how AI-Scientist-v2 optimizes hyperparameters in Stage 2 using a closed feedback loop. Learn about LLM-driven novel proposals and efficient experiment execution to avoid duplicates.

- Tags: how-to-guide
- Published: 2026-03-28

### [The Four Stages of Experiment Execution in AI Scientist v2: A Complete Technical Guide](/SakanaAI/AI-Scientist-v2/stages-of-experiment-execution-ai-scientist-v2)

Explore the four stages of experiment execution in AI Scientist v2 from initial implementation to ablation studies. Learn how AgentManager systematically evolves your research.

- Tags: how-to-guide
- Published: 2026-03-28

### [How AI Scientist v2 Handles Experiment Parallelization: Architecture and Implementation](/SakanaAI/AI-Scientist-v2/how-ai-scientist-v2-handles-experiment-parallelization)

Discover how AI Scientist v2 handles experiment parallelization by spawning Python processes with ParallelAgent and ProcessPoolExecutor for efficient multi-seed evaluations and exclusive GPU allocation.

- Tags: architecture
- Published: 2026-03-28

### [The Role of the Journal in AI Scientist v2 Tree Search: Core Memory Architecture](/SakanaAI/AI-Scientist-v2/role-of-journal-in-ai-scientist-v2-tree-search)

Discover the Journal's crucial role in AI Scientist v2 tree search. Learn how this core memory architecture records, filters, and summarizes every step of the AI's thought process.

- Tags: internals
- Published: 2026-03-28

### [How the Agent Manager Orchestrates Experiments in AI Scientist v2](/SakanaAI/AI-Scientist-v2/how-agent-manager-orchestrates-experiments-ai-scientist-v2)

Discover how the Agent Manager orchestrates experiments in AI Scientist v2. It manages research stages, spawns agents, and drives experiments autonomously based on LLM evaluations.

- Tags: internals
- Published: 2026-03-28

### [How to Configure Tree Search Parameters in AI Scientist v2: 3 Key Settings Explained](/SakanaAI/AI-Scientist-v2/how-to-configure-tree-search-parameters-ai-scientist-v2)

Learn to configure tree search parameters in AI Scientist v2 by adjusting max_debug_depth, debug_prob, and num_drafts in your YAML config for optimized AI research.

- Tags: how-to-guide
- Published: 2026-03-28

### [How to Integrate Custom LLMs with AI Scientist v2: The Complete Developer Guide](/SakanaAI/AI-Scientist-v2/how-to-integrate-custom-llms-ai-scientist-v2)

Learn to integrate custom LLMs with AI Scientist v2. Follow our developer guide to register your model, extend client creation, and adjust request handling for seamless integration.

- Tags: how-to-guide
- Published: 2026-03-28

### [What Are the Dependencies for AI Scientist v2? Complete Package Guide](/SakanaAI/AI-Scientist-v2/what-are-dependencies-ai-scientist-v2)

Discover the dependencies for AI Scientist v2. This guide details the 30+ Python packages needed, including LLM APIs and ML frameworks, all installable with requirements.txt.

- Tags: getting-started
- Published: 2026-03-28

### [How to Set Up the Environment for AI Scientist v2: Complete Installation Guide](/SakanaAI/AI-Scientist-v2/how-to-set-up-environment-ai-scientist-v2)

Install AI Scientist v2 by creating a Python 3.11 Conda environment, setting up CUDA PyTorch, adding PDF/LaTeX tools, and configuring LLM API keys. Complete installation guide.

- Tags: getting-started
- Published: 2026-03-28