Add science ablation labels and batch runner documentation

README.md

@@ -22,6 +22,8 @@ configs:
     features:
       - name: task_id
         dtype: string
+      - name: science_knowledge_ablation
+        dtype: bool
       - name: title
         dtype: string
       - name: domain
@@ -88,6 +90,10 @@ SWE-bench Science evaluates coding agents on software-engineering tasks drawn fr
 | Verifier images | 119 Docker Hub images |
 | Image platform | `linux/amd64` |
 
+The `science_knowledge_ablation` column is `true` for the 91-task science-knowledge
+split used by the ablation experiment. Its release IDs are `002-082`, `084`, `086`,
+`090`, `097-101`, `111`, and `114`; all other rows are `false`.
+
 ## Dataset Viewer And Files
 
 The Dataset Viewer reads the canonical [`data/tasks.csv`](data/tasks.csv) table and generates its preview automatically. The release does not commit a duplicate Parquet export, so the CSV remains the single source of truth for the 119 task rows.
@@ -102,6 +108,7 @@ The repository also includes:
 | `selections/` | Reproducible task selections |
 | `tasks/task_NNN/` | Thin Harbor/Pier task bundles |
 | `tools/` | Materialization, provider, batch, and summary tools |
+| `docs/run-batch.md` | Full provider and batch-runner reference |
 
 The environment image contains the baseline source, public fixtures, dependencies, and compilers. The separate verifier image contains held-out tests and the grader. The dataset does not contain reference-answer patches, credentials, agent trajectories, or private verifier tests.
 
@@ -133,6 +140,16 @@ python3 tools/materialize.py \
   --output tasks-selected-small --force
 ~~~
 
+Materialize the complete 91-task science-knowledge ablation split. It contains
+restricted-license tasks, so the explicit license opt-in is required:
+
+~~~bash
+python3 tools/materialize.py \
+  --task-id 002-082,084,086,090,097-101,111,114 \
+  --allow-restricted-licenses \
+  --output tasks-science-knowledge-ablation --force
+~~~
+
 Every materialization writes `selection.json` with the exact task IDs used for the run.
 
 ## Restricted Licenses
@@ -222,6 +239,10 @@ python3 tools/summarize_results.py --jobs-dir jobs
 
 Use `pier view jobs` to inspect trajectories.
 
+See [`docs/run-batch.md`](docs/run-batch.md) for the complete option reference,
+gateway/profile configuration, dry-run mode, retry and timeout controls, and
+result paths.
+
 ## Licensing And Attribution
 
 The dataset card, release metadata, and helper tools use the repository's MIT terms. Task source, papers, figures, fixtures, and other third-party materials retain their upstream licenses. The source-license field does not automatically license copied scientific materials; audited material notices and modification notes are retained in the relevant task bundles.

data/statistics.md

@@ -4,6 +4,8 @@ Canonical rows: **119**; default unrestricted rows: **96**; restricted-license r
 
 Environment image references: **119/119**; verifier image references: **119/119**.
 
+Science-knowledge ablation rows: **91**.
+
 ### Domain
 
 | Value | Count |

docs/run-batch.md

@@ -0,0 +1,203 @@
+# Batch Evaluation Reference
+
+`tools/run_batch.py` is the convenience wrapper for running a materialized task
+selection with Pier. It does not build task images. It reads the immutable
+environment and verifier references from each `task.toml`, pulls those images
+for `linux/amd64`, writes a redacted run record, and invokes Pier with
+`--no-force-build --no-delete --yes`.
+
+## Prerequisites
+
+From a downloaded release directory:
+
+~~~bash
+uv tool install "datacurve-pier==0.3.0"
+docker login
+python3 tools/materialize.py \
+  --task-id 002,005-007 \
+  --output tasks-selected-small --force
+~~~
+
+The `--path` passed to `run_batch.py` must be a materialized directory containing
+`task_NNN/task.toml` directories. The runner never selects tasks implicitly and
+never reads task definitions from GitHub at runtime.
+
+## Provider Profiles
+
+Use an env file outside the checkout. The parser accepts `KEY=value`, optional
+`export KEY=value`, comments, and quoted values. It never prints credential
+values or writes them to `batch-run.json`.
+
+### Codex and OpenAI-compatible gateways
+
+`run_batch.py` translates the following fields into Pier's Codex provider
+configuration when `--agent codex` is used:
+
+| Variable | Required | Meaning |
+| --- | --- | --- |
+| `MODEL` | No | Exact model route sent to the gateway; default `gpt-5` |
+| `OPENAI_API_KEY` | Yes for a real run | Gateway credential |
+| `CODEX_BASE_URL` | No | OpenAI-compatible gateway URL; defaults to `https://api.openai.com/v1` |
+| `CODEX_WIRE_API` | No | `responses` or `chat`; defaults to `responses` |
+| `CODEX_VERSION` | No | Codex runtime version passed to Pier |
+| `CODEX_REASONING_EFFORT` | No | Reasoning effort passed to the Codex adapter |
+
+Example:
+
+~~~dotenv
+MODEL=gpt-5
+OPENAI_API_KEY=replace-with-your-key
+CODEX_BASE_URL=https://gateway.example.edu/v1
+CODEX_WIRE_API=responses
+CODEX_VERSION=latest
+CODEX_REASONING_EFFORT=high
+~~~
+
+`CODEX_BASE_URL` selects the model gateway. It is different from a network
+proxy. A network proxy is configured with standard `HTTP_PROXY`, `HTTPS_PROXY`,
+and `NO_PROXY` variables. For Docker Desktop, a proxy running on the host is
+usually reached from a container as `host.docker.internal`, not `127.0.0.1`.
+
+### Claude Code and mini-swe-agent
+
+These harnesses receive their provider variables through Pier's `--env-file`:
+
+~~~dotenv
+# Claude Code
+ANTHROPIC_AUTH_TOKEN=replace-with-your-gateway-key
+ANTHROPIC_BASE_URL=https://api.anthropic.com
+ANTHROPIC_CUSTOM_HEADERS=
+~~~
+
+~~~dotenv
+# mini-swe-agent with an OpenAI-compatible provider
+OPENAI_API_KEY=replace-with-your-gateway-key
+OPENAI_BASE_URL=https://gateway.example.edu/v1
+~~~
+
+The model route is selected with the repeatable `--model` option. Provider
+variables not listed here can be added to the env file and are passed through to
+the selected harness by Pier.
+
+## Basic Commands
+
+Run a no-model infrastructure smoke:
+
+~~~bash
+python3 tools/run_batch.py \
+  --path tasks-selected-small \
+  --agent nop \
+  --n-concurrent 1 \
+  --n-attempts 1 \
+  --jobs-dir jobs \
+  --job-name smoke
+~~~
+
+Run Codex through a gateway:
+
+~~~bash
+python3 tools/run_batch.py \
+  --path tasks-selected-small \
+  --agent codex \
+  --env-file ~/.config/swe-bench-science/codex.env \
+  --n-concurrent 2 \
+  --n-attempts 1 \
+  --max-retries 1 \
+  --jobs-dir jobs \
+  --job-name codex-small
+~~~
+
+Run Claude Code or mini-swe-agent:
+
+~~~bash
+python3 tools/run_batch.py \
+  --path tasks-selected-small \
+  --agent claude-code \
+  --env-file ~/.config/swe-bench-science/claude.env \
+  --model anthropic/claude-opus-4-7 \
+  --n-concurrent 1 \
+  --jobs-dir jobs \
+  --job-name claude-small
+~~~
+
+For an approximately 120-second agent-stage smoke, add
+`--agent-timeout-multiplier 0.0223`. This does not shorten the verifier timeout
+or any native build timeout.
+
+## Option Reference
+
+| Option | Default | Description |
+| --- | --- | --- |
+| `--path` | required | Materialized task directory |
+| `--agent` | `nop` | Pier harness, such as `codex`, `claude-code`, `mini-swe-agent`, or `nop` |
+| `--env` | `docker` | Pier environment backend |
+| `--env-file` | unset | Provider/harness env file |
+| `--model` | unset | Model route; repeat for multiple Pier model arguments |
+| `--agent-env KEY=VALUE` | repeatable | Extra environment value passed to the harness |
+| `--agent-kwarg KEY=VALUE` | repeatable | Extra Pier agent keyword; useful for adapter-specific settings |
+| `--n-concurrent` | `1` | Number of simultaneous tasks |
+| `--n-attempts` | `1` | Attempts per task |
+| `--max-retries` | `0` | Pier retries after an attempt-level failure |
+| `--agent-timeout-multiplier` | Pier default | Multiplier for the agent stage timeout |
+| `--verifier-timeout-multiplier` | Pier default | Multiplier for verifier/build timeout |
+| `--jobs-dir` | `jobs` | Directory for Pier jobs and summaries |
+| `--job-name` | unset | Stable job name used in result paths |
+| `--platform` | `linux/amd64` | Docker pull and derived Pier image platform |
+| `--pier-bin` | `pier` | Pier executable or absolute path |
+| `--skip-pull` | off | Skip Docker pulls when immutable refs are already local |
+| `--no-auto-provider` | off | Do not translate `CODEX_*` profile values into Codex kwargs |
+| `--no-auto-agent-adapter` | off | Use Pier's built-in Codex agent instead of the Science Bench adapter |
+| `--agent-import-path` | unset | Explicit Pier agent import path |
+| `--dry-run` | off | Pull/validate images and write metadata, but do not invoke Pier |
+
+The wrapper always records the selected task IDs, selection hash, image refs,
+platform, Pier version, agent/model settings, and a redacted Pier command in
+`<path>/batch-run.json`.
+
+## Results
+
+Pier writes its job output under the selected jobs directory. The wrapper then
+generates:
+
+~~~text
+jobs/<job-name>/result.json
+jobs/<job-name>/summary.json
+jobs/<job-name>/summary.csv
+jobs/<job-name>/<task>__<trial>/verifier/reward.json
+jobs/<job-name>/<task>__<trial>/verifier/ctrf.json
+jobs/<job-name>/<task>__<trial>/verifier/test-stdout.txt
+~~~
+
+The summary CSV is the convenient per-task result table. Use `pier view jobs`
+for trajectories and inspect `result.json`, `reward.json`, and
+`test-stdout.txt` together when diagnosing a failure.
+
+## Common Variants
+
+Pull nothing and inspect the fully rendered command:
+
+~~~bash
+python3 tools/run_batch.py \
+  --path tasks-selected-small \
+  --agent codex \
+  --env-file ~/.config/swe-bench-science/codex.env \
+  --skip-pull \
+  --dry-run
+~~~
+
+Run the 91-task science-knowledge ablation selection after materialization:
+
+~~~bash
+python3 tools/materialize.py \
+  --task-id 002-082,084,086,090,097-101,111,114 \
+  --allow-restricted-licenses \
+  --output tasks-science-knowledge-ablation --force
+
+python3 tools/run_batch.py \
+  --path tasks-science-knowledge-ablation \
+  --agent codex \
+  --env-file ~/.config/swe-bench-science/codex.env \
+  --n-concurrent 4 \
+  --jobs-dir jobs \
+  --job-name codex-science-ablation
+~~~

tasks/task_001/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "001",
+  "science_knowledge_ablation": false,
   "title": "Repair omega-dependent range-separated PBE workflow in a real PySCF STRSF repository",
   "domain": "density-functional-theory",
   "language": "python",

tasks/task_002/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "002",
+  "science_knowledge_ablation": true,
   "title": "Repair an inconsistent reduced validation workflow",
   "domain": "density-functional-theory",
   "language": "python",

tasks/task_003/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "003",
+  "science_knowledge_ablation": true,
   "title": "Repair extxyz stress and virial tensor-label preservation",
   "domain": "atomistic-machine-learning-data",
   "language": "python",

tasks/task_004/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "004",
+  "science_knowledge_ablation": true,
   "title": "Repair split-read exon-overlap filtering",
   "domain": "rna-seq-genomic-interval-semantics",
   "language": "c++",

tasks/task_005/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "005",
+  "science_knowledge_ablation": true,
   "title": "Repair edited MRS subspectrum polarity handling",
   "domain": "magnetic-resonance-spectroscopy-processing",
   "language": "matlab-octave",

tasks/task_006/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "006",
+  "science_knowledge_ablation": true,
   "title": "Repair inconsistent genomic interval enrichment counts",
   "domain": "genomics-interval-statistics",
   "language": "c++",

tasks/task_007/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "007",
+  "science_knowledge_ablation": true,
   "title": "Repair unstable spectral registration for drifted MRS transients",
   "domain": "magnetic-resonance-spectroscopy",
   "language": "python",

tasks/task_008/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "008",
+  "science_knowledge_ablation": true,
   "title": "Establish first vacuum-plus-wall external-region capability in the TERPSICHORE-native path",
   "domain": "plasma-stability",
   "language": "python",

tasks/task_009/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "009",
+  "science_knowledge_ablation": true,
   "title": "Enable projection-wall external-region support in the TERPSICHORE-native path",
   "domain": "plasma-stability",
   "language": "python",

tasks/task_010/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "010",
+  "science_knowledge_ablation": true,
   "title": "Complete alternate perturbation-representation support",
   "domain": "plasma-stability",
   "language": "python",

tasks/task_011/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "011",
+  "science_knowledge_ablation": true,
   "title": "Complete the Missing Response Capability",
   "domain": "plasma-stability",
   "language": "python",

tasks/task_012/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "012",
+  "science_knowledge_ablation": true,
   "title": "Complete 3-D Cross-n FOURIN Spectral Coupling",
   "domain": "plasma-stability",
   "language": "python",

tasks/task_013/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "013",
+  "science_knowledge_ablation": true,
   "title": "Establish boundary-regime operator and energy closure in the TERPSICHORE-native path",
   "domain": "plasma-stability",
   "language": "python",

tasks/task_014/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "014",
+  "science_knowledge_ablation": true,
   "title": "Complete the native multi-equilibrium TERPSICHORE workflow",
   "domain": "plasma-stability",
   "language": "python",

tasks/task_015/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "015",
+  "science_knowledge_ablation": true,
   "title": "Restore thermal-equilibrium invariants in a Liouville-space relaxation workflow",
   "domain": "magnetic-resonance-spin-dynamics",
   "language": "matlab-octave",

tasks/task_016/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "016",
+  "science_knowledge_ablation": true,
   "title": "Repair HMMRATAC peak refinement under sparse signal resolution",
   "domain": "chromatin-accessibility peak calling",
   "language": "python-cython",

tasks/task_017/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "017",
+  "science_knowledge_ablation": true,
   "title": "Repair dark-matter line-of-sight maps in Gammapy",
   "domain": "gamma-ray-astronomy-dark-matter",
   "language": "python",

tasks/task_018/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "018",
+  "science_knowledge_ablation": true,
   "title": "Restore blacklist semantics in an HMMRATAC chromatin-accessibility workflow",
   "domain": "epigenomics-atac-seq-hmm",
   "language": "python-cython",

tasks/task_019/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "019",
+  "science_knowledge_ablation": true,
   "title": "Complete robust topic-to-cistrome binarization for independent probability matrices",
   "domain": "single-cell-chromatin-accessibility",
   "language": "python",

tasks/task_020/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "020",
+  "science_knowledge_ablation": true,
   "title": "Complete an edge-graph geometry and force/virial capability for atomistic models",
   "domain": "atomistic-machine-learning-and-molecular-simulation",
   "language": "python",

tasks/task_021/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "021",
+  "science_knowledge_ablation": true,
   "title": "Complete a missing atomic-liquid thermodynamic-integration capability",
   "domain": "atomistic thermodynamic integration and free-energy calculation",
   "language": "python",

tasks/task_022/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "022",
+  "science_knowledge_ablation": true,
   "title": "Repair an inconsistent volume-to-surface projection",
   "domain": "neuroimaging",
   "language": "python",

tasks/task_023/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "023",
+  "science_knowledge_ablation": true,
   "title": "Restore continuous NPT periodic-boundary unwrapping",
   "domain": "molecular-dynamics-periodic-boundary-analysis",
   "language": "python",

tasks/task_024/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "024",
+  "science_knowledge_ablation": true,
   "title": "Restore coherent inference in a centered factorial-effects model",
   "domain": "probabilistic-scientific-computing",
   "language": "python",

tasks/task_025/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "025",
+  "science_knowledge_ablation": true,
   "title": "Repair inconsistent cross-protocol Osprey validation reports",
   "domain": "edited-magnetic-resonance-spectroscopy-workflow",
   "language": "matlab-octave",

tasks/task_026/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "026",
+  "science_knowledge_ablation": true,
   "title": "Restore isentropic interpolation for terrain-following model data",
   "domain": "atmospheric-science-isentropic-interpolation",
   "language": "python",

tasks/task_027/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "027",
+  "science_knowledge_ablation": true,
   "title": "Restore consistent NMR peak-set encoding under batch padding",
   "domain": "ai4science-nmr-structure-elucidation",
   "language": "python",

tasks/task_028/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "028",
+  "science_knowledge_ablation": true,
   "title": "Repair control-referenced relative responses in a single-cell perturbation workflow",
   "domain": "single-cell-perturbation-response-modeling",
   "language": "python",

tasks/task_029/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "029",
+  "science_knowledge_ablation": true,
   "title": "Restore annotation equivalence after circular-record orientation normalization",
   "domain": "genomics-sequence-annotation",
   "language": "python",

tasks/task_030/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "030",
+  "science_knowledge_ablation": true,
   "title": "Repair format-conformant SHELX/AIRSS .res export in a crystallographic IO workflow",
   "domain": "crystallography-file-format-semantics",
   "language": "python",

tasks/task_031/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "031",
+  "science_knowledge_ablation": true,
   "title": "Restore documented average-reference semantics for mixed-modality recordings",
   "domain": "neurophysiology-signal-processing",
   "language": "python",

tasks/task_032/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "032",
+  "science_knowledge_ablation": true,
   "title": "Repair an inconsistent peptide energy after topology import",
   "domain": "molecular-simulation",
   "language": "python",

tasks/task_033/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "033",
+  "science_knowledge_ablation": true,
   "title": "Repair inconsistent energies across equivalent periodic representations",
   "domain": "periodic-electronic-structure",
   "language": "python",

tasks/task_034/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "034",
+  "science_knowledge_ablation": true,
   "title": "Repair an inconsistent stochastic trajectory calculation",
   "domain": "stochastic-numerical-analysis",
   "language": "python",

tasks/task_035/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "035",
+  "science_knowledge_ablation": true,
   "title": "Repair an incomplete stiff-kinetics campaign",
   "domain": "numerical-analysis-and-chemical-kinetics",
   "language": "python",

tasks/task_036/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "036",
+  "science_knowledge_ablation": true,
   "title": "Repair inconsistent sampling on a spherical curvilinear ocean grid",
   "domain": "computational-oceanography-and-spherical-geometry",
   "language": "python",

tasks/task_037/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "037",
+  "science_knowledge_ablation": true,
   "title": "Repair an inconsistent multiframe MR reconstruction",
   "domain": "medical-imaging",
   "language": "python",

tasks/task_038/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "038",
+  "science_knowledge_ablation": true,
   "title": "Repair inconsistent crystal classification across equivalent representations",
   "domain": "computational-crystallography-and-materials-symmetry",
   "language": "c",

tasks/task_039/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "039",
+  "science_knowledge_ablation": true,
   "title": "Repair a river-network sediment transport calculation",
   "domain": "earth-surface-dynamics-and-river-sediment-transport",
   "language": "python",

tasks/task_040/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "040",
+  "science_knowledge_ablation": true,
   "title": "Restore consistent magnetic-dipole electromagnetic predictions",
   "domain": "computational-electromagnetics-and-geophysical-forward-modeling",
   "language": "python",

tasks/task_041/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "041",
+  "science_knowledge_ablation": true,
   "title": "Repair an inconsistent near-Earth vector conversion",
   "domain": "space-physics-coordinate-systems",
   "language": "python",

tasks/task_042/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "042",
+  "science_knowledge_ablation": true,
   "title": "Repair an inconsistent solar-coordinate archive workflow",
   "domain": "solar-physics-scientific-data-semantics",
   "language": "python",

tasks/task_043/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "043",
+  "science_knowledge_ablation": true,
   "title": "Repair an inconsistent porous-volume characterization",
   "domain": "porous-media-image-analysis",
   "language": "python",

tasks/task_044/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "044",
+  "science_knowledge_ablation": true,
   "title": "Repair inconsistent lithium-ion discharge predictions",
   "domain": "electrochemical-battery-modeling",
   "language": "python",

tasks/task_045/metadata.json

@@ -1,5 +1,6 @@
 {
   "task_id": "045",
+  "science_knowledge_ablation": true,
   "title": "Repair inconsistent finite-volume operations on a connected multi-face grid",
   "domain": "computational-oceanography-finite-volume-grids",
   "language": "python",