Figure 1: Overview of ASI-Bench. Left: B3 performance across agents. Right: scores from B1 to B4.

Overview

ASI-Bench evaluates scientific autonomy: whether an AI agent can translate a scientific objective into an executable and independently validated workflow, rather than merely follow prescribed procedures.

Benchmark scope	Description
Tasks	60 project-level tasks
Scientific coverage	11 domains
Information levels	B1–B4, with progressively less methodological guidance
Evaluation	Expert cross-review, execution testing, and reliable scoring
Evaluated systems	18 agent–model configurations

Across all evaluated systems, the average scores are 50.92 at B1, 29.64 at B2, 27.17 at B3, and 27.33 at B4 as concrete procedural guidance is progressively removed. The sharp drop after B1 and persistently lower scores at B2–B4 show that strong workflow execution does not yet imply independent scientific problem-solving.

Resources

Resource	Link
Project homepage & documentation	asibench.apexin.ai
Source code	apexin-ai/ASI-Bench
Leaderboard	View official results
Dataset · seed42	Apexintelligence-AI/ASI-Bench-seed42
Dataset · seed31415	Apexintelligence-AI/ASI-Bench-seed31415
Python package	asibench on PyPI

Quick start

1. Install ASI-Bench

pip install asibench
asibench --help

Requirements: Python 3.11 or later. The default installation is lightweight and includes everything needed to download public benchmark instances. For local agent execution or task authoring, install the optional full stack:

asibench is the canonical CLI. The legacy ai4sci-bench command remains only as a backwards-compatibility alias.

pip install 'asibench[full]'

2. Download a benchmark set

Each fixed-seed dataset contains its own prompts and input data. Choose one seed for each run:

mkdir asi-bench-run
cd asi-bench-run

asibench task pull \
  --repo seed42 \
  --output-dir hf_instances_seed42/

3. Run an agent

asibench run \
  --instances-dir hf_instances_seed42/ \
  --agent-cmd 'python my_agent.py --workspace {workspace}' \
  --sandbox linux_ns \
  --output-dir out_seed42/

The command above evaluates all four prompt levels (b1,b2,b3,b4). Use --prompt-levels only when intentionally running a subset.

4. Submit for official scoring

asibench login
asibench submit --results-dir out_seed42/

submit creates an authenticated draft on the ASI-Bench website. Review the completeness summary and confirm the draft to enter the official scoring queue. Local benchmark runs do not calculate official scores.

Run configuration

Setting	Behavior
Dataset seed	seed42 and seed31415 are separate runs; use distinct instance and output directories
Prompt levels	All four levels run by default; select a subset with --prompt-levels
Timeout	--timeout defaults to 10,800 seconds and applies uniformly to every task
Submission	submit uploads a draft by default; --no-upload creates a local bundle only
Custom agents	--agent-cmd supports the none and linux_ns sandboxes
Built-in agents	Use --agent with --agent-config; compatible adapters can use Docker-based os isolation

Platform note: The examples use Linux Bash, and the linux_ns sandbox requires Linux. Windows users should run it through WSL2. PowerShell and Command Prompt use different line-continuation and JSON-escaping syntax.

See Getting Started for agent configuration, sandbox selection, and platform-specific commands.

Package and CLI compatibility

• asibench is the only published Python distribution and the canonical CLI.
• Existing Python integrations may continue to import ai4sci_bench.
• ai4sci-bench remains available as a legacy CLI alias.
• Benchmark instances are downloaded separately with asibench task pull.
• Runtime and declared-output metadata come from tasks/ or --tasks-dir.
• Retired internal orchestration commands are not part of the public CLI. Use run, report, batch-report, review, and the Portal workflows.

Built-in CLI agents — installation

Install the CLI agent you want to evaluate, then select its ASI-Bench adapter with --agent.

• New model, existing harness. When a supported harness such as Codex CLI, Claude Code, or Kimi Code adds a model, select the new model through --agent-config. If the model is served through a third-party endpoint and the adapter supports that route, also set api_base, api_key, and api_protocol (openai or anthropic). No ASI-Bench code change is normally needed while the harness accepts the model and the endpoint remains protocol compatible.
• New agent harness. A new CLI or runtime can be used immediately through --agent-cmd when it can read the prepared workspace and write result files; this generic mode supports none and linux_ns. First-class built-in integration—its own --agent name, harness-specific authentication and configuration, proxy wiring, and Docker-based os support—requires ASI-Bench maintainers to implement and test a new adapter.

Agent	Install command	Notes
Claude Code	npm install -g @anthropic-ai/claude-code	Requires Anthropic API key or local login (claude /login)
Codex CLI	npm install -g @openai/codex	Requires OpenAI API key
Kimi Code	npm install -g @moonshot-ai/kimi-code	Use the npm package. The PyPI package (kimi-cli) has different CLI arguments and is not compatible with this adapter. Requires Moonshot API key or local login (kimi /login)
CodeWhale	See CodeWhale docs	Requires DeepSeek API key by default
AntiGravity	npm install -g @anthropic-ai/antigravity	Experimental

Example with Kimi Code:

asibench run --agent kimi_code_cli \
  --agent-config '{"model": "kimi-k2.7"}' \
  --output-dir out/

Framework-managed model proxies use text-only input by default. For endpoints that accept images, set "supports_image_input": true in --agent-config. Without this flag, image attachments are replaced with a short notice before the next model request. Direct endpoints outside the framework proxy remain the responsibility of the connected agent.

Contribute a task

We welcome project-level scientific tasks from the research community. Authors can draft a task locally or in the Portal, validate it, and submit a frozen revision for review.

Quick start for task authors

# 1. Create a task scaffold
asibench task create --domain physics --name my_new_task

# 2. Edit the generated files in tasks/physics/my_new_task/:
#    - task_meta.yaml   — task metadata (public)
#    - task_eval.yaml   — evaluation config: scoring rules, parameter space (private)
#    - generate_gt.py   — ground-truth generator (private)
#    - custom_scorer.py — custom scoring logic, if needed (private)
#    - prompt_b1.md … prompt_b4.md — difficulty-ladder prompts (public)

# 3. Required pre-submit validation
asibench validate --pre-submit tasks/physics/my_new_task/

# 4. Required same-model trial across B1, B2, B3, and B4
asibench difficulty-check --task physics.my_new_task

# 5. Open the Portal task-submission form
asibench task submit

Submission requirements

• Complete validate --pre-submit and a same-model difficulty-check across B1–B4 before submission.
• Use lowercase letters, digits, and underscores for task IDs. The canonical form is domain.task_name.
• asibench task submit opens the Portal form; files are uploaded and reviewed in the browser, not from the terminal.
• Set ASIBENCH_NO_BROWSER=1 to print the form URL without opening a browser.
• The Portal requires local_testing_done and one finite score for each level before freezing a revision.
• Authors do not create repository pull requests. Administrators publish accepted tasks after review.

The default submission form is asibench.apexin.ai/submit/proposals/new. Use --endpoint or ASIBENCH_SUBMIT_ENDPOINT to select another Portal.

Runnable samples from the official benchmark

The following final tasks can be used to verify the runner workflow:

Task	Domain	Description
astronomy.nbody_close_encounters	Astronomy	Close-encounter few-body scattering
math.homotopy_poly_roots	Math	Isolated complex roots via homotopy continuation

Pull and run one in produce-only mode:

asibench task pull \
  --repo seed42 \
  --tasks astronomy.nbody_close_encounters \
  --output-dir example_instances/

asibench run \
  --instances-dir example_instances/ \
  --tasks astronomy.nbody_close_encounters \
  --agent-cmd 'python my_agent.py --workspace {workspace}' \
  --sandbox linux_ns \
  --output-dir example_results/

asibench submit --results-dir example_results/

For official benchmark tasks, the GitHub catalog contains metadata only. Prompts and input data come from Hugging Face, while reference answers and scorers remain on the private scoring service.

Five opt-in sample tasks are fully public examples. Their B1–B4 prompts, ground-truth generators, and scorer implementations or configurations are available under tasks/:

• robotics.minimum_snap_trajectory_conditioning
• chemistry.bsse_counterpoise_cbs_extrapolation
• materials.phonon_dispersion
• medicine.ethics_disclosure_diagnosis_1670
• medicine.jama_id0014_malignant_an

The examples were synchronized from apexin-ai/Agent-AI4Sci-Bench at revision 2ba9258442bf53ad6c4911957234e03e767476ad.

Dataset layout

Item	Contract
Official tasks	60 final task definitions shared by both fixed-seed datasets
Public examples	5 opt-in sample tasks with complete prompts and scoring assets
Instance layout	<output-dir>/<instance-id>/ after asibench task pull
Default selection	Unfiltered official runs exclude sample tasks
Including samples	Use --include-sample or name a sample explicitly with --tasks

Pull the two fixed-seed datasets separately:

asibench task pull --repo seed42 --output-dir hf_instances_seed42/
asibench task pull --repo seed31415 --output-dir hf_instances_seed31415/

seed42 and seed31415 are the only official Hugging Face contracts, and --repo is required. They share task schemas and declared outputs but contain different generated inputs. Matching schemas live in tasks/**/task_meta.yaml; private manifests, references, and scorers stay on the scoring side.

The task workflow uses the grouped asibench task create, task pull, and task submit commands. See Contribute a Task for the complete authoring guide.

How scoring works (and stays fair)

• Public execution framework: agent adapters, sandboxes, instance loading, output collection, and submission packaging are auditable in this repository.
• Centralized official scoring: asibench login identifies the submitter, and asibench submit sends a draft for confirmation and scoring. Self-reported scores are not accepted.

Sandboxing & reproducibility

• Runs support task, os, and linux_ns sandboxes with pinned per-task runtimes.
• Every run records agent, model, effort, sandbox, and framework provenance.
• Source checkouts are installed editable into task environments; wheel-based runs install the matching asibench version.
• Mutable caches live under ~/.asibench/, never under site-packages.
• Namespace and Docker prerequisites are checked before execution. Missing prerequisites produce a non-zero CLI exit.

Documentation

Installation, execution, submission, task contribution, catalog, and leaderboard documentation are available on the ASI-Bench website.

License

See LICENSE.