Update README.md

README.md

@@ -73,215 +73,303 @@ configs:
   - split: test
     path: data/core/geodisaster_2921_test_source.jsonl
 ---
+---
 
-# GeoDisaster Agentic 2921
+<div align="center">
 
-Ground-truth **agentic tool-use chains** for geospatial disaster analysis: **2,921 tasks**, each a
-question grounded in real Earth-observation artifacts (optical/SAR imagery, forest masks, hazard
-rasters, building-damage polygons, routing graphs), solved end-to-end by a tool-using agent — in
-**two settings**:
+# 🌍 GeoDisaster
 
-- **single-agent** — one agent plans + calls tools itself.
-- **multi-agent** — orchestrator **ORC** delegates to specialists **GA / PA / VRA**.
+### Benchmarking Orchestrated Agents for Operational Disaster Geo-Intelligence
 
-**2,921 tasks** (train 2,629 / test 292) · **43 question types** · 5 families:
-`deforestation`, `multi_hazard`, `xbd_damage`, `routing`, `sen1_flood`.
+**Accepted at BMVC 2026**
 
-## IDs & paths
+<a href="https://arxiv.org/pdf/2606.17246"><img alt="Paper" src="https://img.shields.io/badge/Paper-arXiv%3A2606.17246-B31B1B?style=flat-square" /></a>
+<a href="https://github.com/VIMAGE-IITB/GeoDisaster"><img alt="GitHub" src="https://img.shields.io/badge/Code-GitHub-24292e?style=flat-square&logo=github" /></a>
+<a href="https://huggingface.co/datasets/vimageiitb/geodisaster-agentic-2921"><img alt="Hugging Face" src="https://img.shields.io/badge/Dataset-Hugging%20Face-FFD21E?style=flat-square" /></a>
+<a href="https://arxiv.org/abs/2606.17246"><img alt="BMVC 2026" src="https://img.shields.io/badge/BMVC-2026%20Accepted-2E8B57?style=flat-square" /></a>
 
-Each record has `id` = `"<family>/<sample>/<question_type>"` (e.g.
-`deforestation/000086/multi_year_loss_attribution`) — the key that links the same task across all
-forms — plus `sample_id` and `question_type`. Artifacts live at `source_assets/<family>/<sample>/…`.
+**[📄 Paper](https://arxiv.org/pdf/2606.17246)** · **[💻 Code & documentation](https://github.com/VIMAGE-IITB/GeoDisaster)** · **[🤗 Dataset](https://huggingface.co/datasets/vimageiitb/geodisaster-agentic-2921)**
 
-## Tool ownership
+*Grounded, multi-step disaster geo-intelligence from real Earth-observation and GIS evidence.*
 
-| Role | Tools |
-|---|---|
-| **ORC** | plans & delegates (no tools) |
-| **GA** | ReadData, PrepareScene, MeasureArea, ExtractObjects, MapFlood, RenderMap |
-| **PA** | ComputeStats, CompareSources, DetectChange, ClassifyChange, ScoreHazard, EstimateExposure, RankRegions, PlanRoutes, CompareRoutes |
-| **VRA** | AssessDamage, DescribeFlood, CompareScenes |
+</div>
+
+---
+
+## ✨ Benchmark at a glance
+
+| **2,921** verified tasks | **43** question types | **5** disaster families | **2** agent settings |
+|:---:|:---:|:---:|:---:|
+| **2,629** train · **292** test | **55,648** turn-level SFT rows | **18** geospatial tools | Optical · SAR · Raster · Vector · Graph |
+
+**GeoDisaster** evaluates whether tool-using agents can transform heterogeneous Earth-observation (EO) and Geographic Information System (GIS) inputs into **structured, evidence-grounded disaster decisions**. Each task connects a real geospatial question to declared input artifacts, an executable reference toolchain, and a verifiable answer target.
+
+The dataset provides matched records for **single-agent** reasoning and **orchestrated multi-agent** execution. The same **2,921 underlying tasks** are available as reference task specifications, full conversations, and turn-level training examples; these are different representations of the same benchmark, *not* additional independent tasks.
+
+> **For evaluation:** use the gold-free `evaluation_inputs` configuration / `data/core/geodisaster_2921_test_inputs.jsonl`. Reference answers and test trajectories must remain on the evaluator side. See [Evaluation without answer leakage](#-evaluation-without-answer-leakage).
+
+## 🌋 Five disaster task families
+
+| Task family | Dataset key | What it evaluates |
+|---|---|---|
+| 🌲 **Deforestation Monitoring** | `deforestation` | Forest-loss detection, multi-year attribution, and change measurement from EO layers |
+| 🌪️ **Multi-Hazard Analysis** | `multi_hazard` | Hazard statistics, cross-layer comparison, spatial exposure, and region prioritization |
+| 🏚️ **Building-Damage Assessment** | `xbd_damage` | Damage assessment and inventory from building footprints and disaster evidence |
+| 🛣️ **Flood-Safe Routing** | `routing` | Network reasoning, hazard-aware route planning, and route comparison |
+| 🌊 **Sentinel-1 SAR Flood Monitoring** | `sen1_flood` | SAR-driven flood mapping, flood measurement, and structured flood reporting |
+
+**Evidence modalities:** optical and SAR imagery; forest-loss masks; hazard and flood rasters; building-damage polygons; vector geometries; road networks; exposure layers; and structured geospatial artifacts.
+
+## 📦 What is included?
+
+| Representation | Purpose | Contents |
+|---|---|---|
+| **Core benchmark** | Evaluation and reference | Task/question specification, available artifact handles, canonical ground-truth toolchain, structured expected answer |
+| **Full conversations** | Trajectory analysis and supervised examples | Complete agent episodes: question → tool calls / observations → terminal answer, for single- and multi-agent settings |
+| **Turn-level SFT** | Training and role-specific analysis | One assistant target per row, with the required context and metadata |
+
+| Release statistics | Count |
+|---|---:|
+| Unique benchmark tasks | **2,921** |
+| Training tasks (per full agent form) | **2,629** |
+| Test tasks (per full agent form) | **292** |
+| Single-agent turn-level rows | **20,453** |
+| Multi-agent role-level rows | **35,195** |
+| **Total turn-level rows** | **55,648** |
+| Terminal targets across both full agent forms | **5,842** |
 
-## Files & forms
+Training and test partitions are disjoint by **task ID, source sample, declared path, normalized routing content, and complete model-visible single-agent trajectory**. Within a split, exact repeated episodes are weighted by inverse complete-trajectory multiplicity via `training_weight`.
 
+## 🚀 Quick start
+
+Load the released configurations with the Hugging Face `datasets` library:
+
+```python
+from datasets import load_dataset
+
+REPO = "vimageiitb/geodisaster-agentic-2921"
+
+core        = load_dataset(REPO, "core")
+eval_inputs = load_dataset(REPO, "evaluation_inputs")
+
+single      = load_dataset(REPO, "single_agent")
+multi       = load_dataset(REPO, "multi_agent")
+
+single_sft  = load_dataset(REPO, "single_agent_sft")
+ga_sft      = load_dataset(REPO, "multi_agent_sft_ga")
 ```
+
+Other specialist role-SFT files are available in the dataset repository under `data/multi_agent/sft_train/` and `data/multi_agent/sft_test/`. You can also read any released `.jsonl` file directly, one record per line.
+
+**Choosing a view:** use `core` for reference specifications, `single_agent` / `multi_agent` for full episodes, SFT views for assistant-turn supervision, and **`evaluation_inputs` for agent-visible held-out tasks**. Never give reference-bearing test files to the evaluated agent.
+
+## 🗂️ Repository layout
+
+```text
 data/
-  core/        geodisaster_2921_{train,test}_source.jsonl   # evaluator/reference task specs
-               geodisaster_2921_test_inputs.jsonl           # gold-free agent-visible test view
-  single_agent/
-    {train_conversations,test_reference_conversations}.jsonl  # full episodes
-    {train_turns,test_turns}.jsonl                            # per-turn SFT rows
-    single_agent_system_prompt.md
-  multi_agent/
-    {train,test}_conversations.jsonl                          # full ORC↔GA/PA/VRA episodes
-    sft_{train,test}/{orc,ga,pa,vra}_{train,eval}.jsonl       # weighted per-role SFT rows
-    tool_ownership_and_prompt_context.md
-source_assets/   bundled inputs; pinned installer hydrates Sen1Floods11 execution inputs
-user-guide-final.md  authoritative benchmark, runtime, release, and 43-type guide
-ATTRIBUTIONS.md  upstream sources & licenses
-SHA256SUMS      integrity manifest for the distributable thin archive
+├── core/
+│   ├── geodisaster_2921_{train,test}_source.jsonl   # evaluator/reference tasks
+│   ├── geodisaster_2921_test_inputs.jsonl          # gold-free public test inputs
+│   └── scoring_policy_43.json                     # active structured-answer policy
+├── single_agent/
+│   ├── {train_conversations,test_reference_conversations}.jsonl
+│   ├── {train_turns,test_turns}.jsonl
+│   └── single_agent_system_prompt.md
+└── multi_agent/
+    ├── {train,test}_conversations.jsonl
+    ├── sft_train/{orc,ga,pa,vra}_{train,eval}.jsonl
+    ├── sft_test/{orc,ga,pa,vra}_{train,eval}.jsonl
+    └── tool_ownership_and_prompt_context.md
+
+source_assets/                                  # bundled permissible EO/GIS inputs
+release_tools/
+├── fetch_sen1floods11_subset.py                # pinned SAR input installer
+└── build_isolated_test_bundle.py               # held-out evaluation bundle
+user-guide-final.md                             # authoritative benchmark/runtime guide
+ATTRIBUTIONS.md                                # upstream licensing and provenance
+SHA256SUMS                                     # distributable archive checksums
 ```
 
-`SHA256SUMS` deliberately excludes locally hydrated
-`source_assets/sen1_flood/` files. Those 300 files are verified against the
-pinned upstream manifest by the hydrator itself.
+<details>
+<summary><b>🔍 Record IDs, shared fields, and dataset schema</b></summary>
+
+### Record identity
 
-Read `user-guide-final.md` for the complete current contract. It is the sole
-task, interpretation, runtime, and release guide included in the package.
+Every representation shares three identity fields:
 
-The same 2,921 tasks come in three forms: **core** (task spec + GT toolchain + answer),
-**conversation** (full episode: question → turns → answer), **turns** (the episode sliced to one
-assistant target per row, for SFT). All files are JSONL (one record per line).
+- `id`: `<family>/<sample>/<question_type>` — the unique linking key across benchmark forms.
+- `sample_id`: source sample identifier.
+- `question_type`: one of the 43 question types.
 
-The release contains 2,629 train and 292 test tasks in each full agent form. Its turn-level
-training projections contain 35,195 multi-agent role rows and 20,453 single-agent rows (55,648
-total). All 5,842 terminal targets use the same reference-only action contract.
+For example, `deforestation/000086/multi_year_loss_attribution` identifies the same task across its core specification, conversations, and SFT turns. Source input files use repository-relative paths such as `source_assets/deforestation/000086/masks/lossyear.tif`. Non-distributed teacher replay outputs, if referenced, use portable `generated-artifact://` logical locators.
 
-The only multi-agent role-SFT locations are `data/multi_agent/sft_train/` and
-`data/multi_agent/sft_test/`. Every conversation and SFT row carries the shared
-`id`, `sample_id`, and `question_type` identity fields. Packaged inputs use
-repository-relative paths; non-distributed replay outputs, when mentioned in
-teacher evidence, use portable `generated-artifact://` logical locators.
+### Multi-agent conversation fields
 
-In the 500 Sen1Floods11 records, `exists: true` describes the hydrated
-execution contract. The thin public archive intentionally lacks the exact 200
-manifest-declared S1/JRC source paths until the pinned installer downloads and
-verifies them and generates the 100 portable execution rasters.
+Files: `data/multi_agent/{train,test}_conversations.jsonl`
 
-## Dataset structure (fields per record)
+| Field | Type | Description |
+|---|---|---|
+| `raw_question` | string | Plain question used for active orchestrator/specialist prompts |
+| `question` | string | Stored evaluation/display context: question, input handles, and answer schema; active prompts use `raw_question` with sanitized handle cards |
+| `initial_artifacts` | object | Declared input handle cards with path, label, existence, and optionality |
+| `available_initial_handles` | list | Handles available to the task |
+| `plan` | object | ORC plan with goal and `{id, agent, subtask, needs}` steps |
+| `multiagentic_gt_chain` | list | Full USER ↔ ORC ↔ specialist trajectory, actions, and observations |
+| `executions` | list | Per-step `{action, tool_output}` execution log |
+| `expected_final_answer` | object | Audit copy of the reference answer; the core record is authoritative |
+| `final_answer_ref` | string | Terminal transport token (`ANSWER_JSON:answer_1`) |
+| `resolved_final_answer` | object | Schema-valid answer projected from worker evidence |
+| `resolved_answer_sha256` | string | Canonical-JSON checksum linking derived records |
+| `resolved_answer_provenance` | string | Export provenance: full evidence or validated compact-trace cache |
+| `answer_support` | object | Supporting evidence metadata |
+| `pure_toolchain_sim_status` | string | Toolchain replay status |
+
+### Other released record types
+
+| Path | Main fields |
+|---|---|
+| `data/core/*_source.jsonl` | `question`, `initial_artifacts`, `available_initial_handles`, `gt_chain`, `expected_final_answer`, `source_replay`, `path_summary`, `split_seed` |
+| `data/single_agent/{train_conversations,test_reference_conversations}.jsonl` | `messages` (`role`, `content`), `tool_sequence`, `final_answer_ref`, `resolved_answer_sha256` |
+| `data/single_agent/{train_turns,test_turns}.jsonl` | `messages`, `kind`, `assistant_action`, `assistant_turn_index` |
+| `data/multi_agent/sft_{train,test}/*.jsonl` | `messages`, `role`, `step_id`, `turn_no`, `kind` |
 
-**Shared (all files):** `id` (string, unique linking key) · `sample_id` (string) ·
-`question_type` (string).
+Each role-SFT row contains the exact system-prompt snapshot associated with its target. The data package also includes the standalone single-agent prompt and the multi-agent role/tool prompt context. **Executable inference-time prompt builders are authoritative in the [code repository](https://github.com/VIMAGE-IITB/GeoDisaster).**
 
-**`data/multi_agent/{train,test}_conversations.jsonl`** — one full episode:
+</details>
 
-| Field | Type | Meaning |
-|---|---|---|
-| `raw_question` | string | plain question used to construct active ORC/specialist prompts |
-| `question` | string | stored evaluator/display context: question + declared input handles + answer schema; executable prompts use `raw_question` plus sanitized handle cards and keep the schema runtime-side |
-| `initial_artifacts` | object | handle cards `{path,label,exists,optional}`; lists each file once |
-| `available_initial_handles` | list | handle names the task exposes |
-| `plan` | object | ORC plan `{goal, steps:[{id,agent,subtask,needs}]}` |
-| `multiagentic_gt_chain` | list | turn-by-turn episode (USER↔ORC↔GA/PA/VRA) with actions + observations |
-| `executions` | list | per-step `{action, tool_output}` log |
-| `expected_final_answer` | object | canonical audit copy; core JSONL is authoritative |
-| `final_answer_ref` | string | terminal transport token, always `ANSWER_JSON:answer_1` |
-| `resolved_final_answer` | object | schema-valid worker projection resolved by the runtime |
-| `resolved_answer_sha256` | string | canonical-JSON checksum linking derived exports |
-| `resolved_answer_provenance` | string | whether the export retained full worker evidence or used its validated compact-trace cache |
-| `answer_support` | object | supporting evidence metadata |
-| `pure_toolchain_sim_status` | string | replay status |
-
-**`data/single_agent/{train_conversations,test_reference_conversations}.jsonl`:** `messages`
-(chat turns `{role,content}`), `tool_sequence` (list), `final_answer_ref`, and
-`resolved_answer_sha256` linking the corresponding verified multi-agent payload.
-
-**`data/single_agent/{train_turns,test_turns}.jsonl`** (SFT): `messages` (context → one assistant
-target), `kind`, `assistant_action`, `assistant_turn_index`.
-
-**`data/multi_agent/sft_{train,test}/...jsonl`** (SFT): `messages`, `role`, `step_id`,
-`turn_no`, `kind`.
-
-**`data/core/..._source.jsonl`:** `question` (obj), `initial_artifacts`, `available_initial_handles`,
-`gt_chain` (canonical tool sequence), `expected_final_answer`, `source_replay`, `path_summary`,
-`split_seed`.
-
-**Splits:** train 2,629 / test 292 per form. They are disjoint by `id`, source
-sample, declared path, normalized routing content, and exact complete
-model-visible single-agent trajectory. Exact episodes retained within one
-split use inverse complete-trajectory multiplicity in `training_weight`.
-
-## Public-reference evaluation isolation
-
-Use `data/core/geodisaster_2921_test_inputs.jsonl` as the only test-record file
-visible to an evaluated agent. The canonical, `test_source`, multi-agent test
-conversations, and single-agent test-reference files contain evaluator targets
-and must remain evaluator-side.
-
-For filesystem isolation, build a fresh root containing only the 292 test
-records and their permitted inputs:
+## 🛡️ Evaluation without answer leakage
+
+**Only `data/core/geodisaster_2921_test_inputs.jsonl` should be exposed as test records to an evaluated agent.** The canonical reference data, `test_source`, multi-agent test conversations, and single-agent test-reference conversations contain evaluator-only targets.
+
+For a filesystem-isolated evaluation, create a separate root containing just the **292 held-out test tasks** and their permitted inputs:
 
 ```bash
-python3 release_tools/build_isolated_test_bundle.py --output /path/outside/repository/eval_bundle
+python3 release_tools/build_isolated_test_bundle.py \
+  --output /path/outside/repository/eval_bundle
+
 export GEODISASTER_DATASET_ROOT=/path/outside/repository/eval_bundle
 export GEODISASTER_STRICT_DATA_ROOT=1
 ```
 
-The builder rejects private path declarations, omits `gt/`, `ground_truth/`,
-`hidden/`, and trajectory products when a declared input is a directory, and
-writes a SHA-256 file manifest. Strict runtime mode disables fallback searches
-through the development repository. Keep gold scoring in a separate evaluator
-process or mount; never mount the repository root into the agent/tool sandbox.
+The bundle builder rejects private path declarations; omits `gt/`, `ground_truth/`, `hidden/`, and trajectory products when a declared input is a directory; and writes a SHA-256 manifest. Strict runtime mode prevents fallback searches through the development repository. **Keep gold scoring in a separate process or mount; do not mount the full dataset/development repository into the agent sandbox.**
+
+The active structured-answer scoring policy is `data/core/scoring_policy_43.json` (`geodisaster_structured_answer_v2`). Older v1 policy text inside `correction_provenance.publication_contract` is historical publication provenance, **not** the active evaluation policy. Values in `correction_provenance.evidence` are opaque build labels, not runtime paths or agent inputs.
+
+## 🧑‍🤝‍🧑 Agent settings and tool ownership
+
+The released trajectories cover two settings:
+
+- **Single-agent:** one agent plans, invokes tools, and finishes the task.
+- **Multi-agent:** an orchestrator (**ORC**) assigns work to **GA**, **PA**, and **VRA** specialists.
+
+| Role | Tools / responsibility |
+|---|---|
+| **ORC** — Orchestrator | Planning and delegation; owns no geospatial tools |
+| **GA** — Geospatial Agent | `ReadData`, `PrepareScene`, `MeasureArea`, `ExtractObjects`, `MapFlood`, `RenderMap` |
+| **PA** — Planning/Analysis Agent | `ComputeStats`, `CompareSources`, `DetectChange`, `ClassifyChange`, `ScoreHazard`, `EstimateExposure`, `RankRegions`, `PlanRoutes`, `CompareRoutes` |
+| **VRA** — Visual Reasoning Agent | `AssessDamage`, `DescribeFlood`, `CompareScenes` |
+
+**Terminal-answer contract:** all **5,842** full-episode terminal targets use the reference-only token `ANSWER_JSON:answer_1`. The runtime creates `answer_1` only from a schema-valid projection of worker evidence, resolves it at termination, and passes the resulting JSON to the evaluator. The token is **transport, not a second answer format**; agents do not emit the large resolved JSON themselves. Unknown or missing references fail explicitly, with no ground-truth fallback in live execution.
 
-## Example instances
+- Multi-agent ORC ends with `Terminate`; the single-agent baseline ends with `FinalAnswer`. Both carry the same answer reference.
+- `TaskComplete` closes a specialist subgoal; it does **not** transmit the terminal answer.
+- Single-agent training support covers SFT and deterministic/live evaluation; **GRPO/RCEA training is supported only for the four multi-agent roles**.
+
+<details>
+<summary><b>🧪 Example task and trajectory records</b></summary>
+
+**Core task specification** (abridged for readability):
 
-Core (task spec):
 ```json
-{"id":"deforestation/000086/multi_year_loss_attribution","sample_id":"000086",
- "question_type":"multi_year_loss_attribution",
- "question":{"handle":"question_1","text":"… identify the year with the largest forest loss …"},
- "initial_artifacts":{"forest_loss_history_1":{"handle":"forest_loss_history_1","type":"forest_loss_history","paths":{"…":[{"path":"source_assets/deforestation/000086/masks/lossyear.tif","label":"lossyear.tif","exists":true,"optional":false}]}}},
- "gt_chain":[{"tool":"ReadData","args":{"input":"forest_loss_history_1"}},{"tool":"ComputeStats","args":{}}],
- "expected_final_answer":{"year":2016,"loss_ha":201.039833}}
+{
+  "id": "deforestation/000086/multi_year_loss_attribution",
+  "sample_id": "000086",
+  "question_type": "multi_year_loss_attribution",
+  "question": {"handle": "question_1", "text": "… identify the year with the largest forest loss …"},
+  "gt_chain": [
+    {"tool": "ReadData", "args": {"input": "forest_loss_history_1"}},
+    {"tool": "ComputeStats", "args": {}}
+  ],
+  "expected_final_answer": {"year": 2016, "loss_ha": 201.039833}
+}
 ```
 
-Single-agent conversation:
+This example's `initial_artifacts` include a declared `forest_loss_history_1` handle pointing to `source_assets/deforestation/000086/masks/lossyear.tif`.
+
+**Single-agent conversation** (abridged; the stored record includes full messages):
+
 ```json
-{"id":"deforestation/000086/multi_year_loss_attribution","sample_id":"000086","question_type":"multi_year_loss_attribution",
- "tool_sequence":["ReadData","PrepareScene","ComputeStats","DetectChange"],
- "messages":[{"role":"system","content":"You are a single geospatial-disaster tool-use agent. …"},
-   {"role":"user","content":"QUESTION …\n\nAVAILABLE INPUT HANDLES\n- forest_loss_history_1: …"},
-   {"role":"assistant","content":"{\"thought\":\"…\",\"action\":{\"tool\":\"ReadData\",\"args\":{\"input\":\"forest_loss_history_1\"}}}"},
-   {"role":"assistant","content":"{\"thought\":\"…\",\"action\":{\"tool\":\"FinalAnswer\",\"answer_ref\":\"ANSWER_JSON:answer_1\"}}"}]}
+{
+  "id": "deforestation/000086/multi_year_loss_attribution",
+  "tool_sequence": ["ReadData", "PrepareScene", "ComputeStats", "DetectChange"],
+  "messages": [
+    {"role": "user", "content": "QUESTION … AVAILABLE INPUT HANDLES …"},
+    {"role": "assistant", "content": "{\"action\":{\"tool\":\"ReadData\",\"args\":{\"input\":\"forest_loss_history_1\"}}"},
+    {"role": "assistant", "content": "{\"action\":{\"tool\":\"FinalAnswer\",\"answer_ref\":\"ANSWER_JSON:answer_1\"}}"}
+  ]
+}
 ```
 
-Multi-agent: same `id`, plus `plan`, `multiagentic_gt_chain` (USER/ORC/GA/PA/VRA turns), `executions`,
-`final_answer_ref`, `resolved_final_answer`, and `expected_final_answer`. Every role-SFT row carries
-the exact system-prompt snapshot used for that target. The standalone single-agent prompt and the
-multi-agent role/tool context are also included with the data; executable prompt builders live in
-the code repository and are authoritative at inference.
-
-The reference is transport, not a second answer format. Agents never reproduce large answer JSON.
-The runtime owns `answer_1`, creates it only from a schema-valid projection of worker evidence,
-resolves it at termination, and passes the resolved JSON to the strict evaluator. Missing or unknown
-references fail explicitly; there is no gold-answer or best-effort fallback during live execution.
-`TaskComplete` closes one specialist subgoal and never carries the final answer. Multi-agent ORC
-uses `Terminate`; the single-agent baseline uses `FinalAnswer`; both carry exactly
-`ANSWER_JSON:answer_1`.
-
-The active answer policy is `data/core/scoring_policy_43.json`
-(`geodisaster_structured_answer_v2`). A v1 policy name nested under
-`correction_provenance.publication_contract` records the historical publication
-pass and does not configure evaluation. Values under
-`correction_provenance.evidence` are opaque build-run provenance labels, not
-runtime filesystem paths and not agent inputs. Single-agent learning consists of SFT
-and deterministic/live evaluation; GRPO/RCEA is supported only for the four
-multi-agent roles.
-
-## Load
+The corresponding multi-agent example uses the same task `id`, and additionally includes `plan`, `multiagentic_gt_chain`, `executions`, `resolved_final_answer`, and reference-answer metadata.
 
-```python
-from datasets import load_dataset
-multi  = load_dataset("vimageiitb/geodisaster-agentic-2921", "multi_agent")
-single = load_dataset("vimageiitb/geodisaster-agentic-2921", "single_agent")
-single_sft = load_dataset("vimageiitb/geodisaster-agentic-2921", "single_agent_sft")
-ga_sft = load_dataset("vimageiitb/geodisaster-agentic-2921", "multi_agent_sft_ga")
-core   = load_dataset("vimageiitb/geodisaster-agentic-2921", "core")
-eval_inputs = load_dataset("vimageiitb/geodisaster-agentic-2921", "evaluation_inputs")
-```
-Or read any JSONL line-by-line.
+</details>
+
+## 🛰️ Source assets and Sen1Floods11 setup
 
-## Sources & license
+Bundled files in `source_assets/` cover redistributable EO/GIS inputs, while the selected **Sen1Floods11** S1/JRC payload is intentionally **not included** in the thin public release.
 
-Project-authored annotations/questions/chains: **CC-BY-4.0**. This repository as
-a whole is **mixed-license**, so the dataset-card license is `other`. Bundled
-`source_assets/` retain their upstream terms. xBD is attributed to the original
-xBD/xView2 sources and remains non-commercial. The public package excludes the
-Sen1Floods11 payload; users hydrate the pinned 100-chip S1/JRC subset directly
-from the official bucket with:
+To download the pinned **100-chip subset** from the official upstream bucket and verify the required files:
 
 ```bash
 python3 release_tools/fetch_sen1floods11_subset.py
 python3 release_tools/fetch_sen1floods11_subset.py --verify-only
 ```
 
+**Important reproducibility details:**
+
+- The **500 `sen1_flood` task records** describe a hydrated execution contract; `exists: true` does not mean every source file is already present in the downloaded archive.
+- The installer fetches and verifies **200 upstream S1/JRC source paths** and produces **100 portable execution rasters** (300 hydrated files in total).
+- `SHA256SUMS` covers the distributed thin archive but **excludes** locally hydrated files under `source_assets/sen1_flood/`; the installer verifies those separately using its pinned upstream manifest.
+- Dataset input paths are repository-relative. Follow `user-guide-final.md` for task-specific requirements and correct execution semantics.
+
+## ⚖️ Licensing and attribution
+
+**This is a mixed-license dataset.** Its Hugging Face license field is therefore `other`, not a blanket CC-BY-4.0 declaration.
+
+| Component | Applicable terms |
+|---|---|
+| Project-authored questions, annotations, and tool-use chains | **CC BY 4.0** |
+| Redistributed `source_assets/` | Retain their respective upstream licenses and conditions |
+| xBD / xView2-derived assets | **Non-commercial restrictions** from their original sources |
+| Selected Sen1Floods11 subset | Retrieved from the official source; subject to upstream terms |
+
+Read **`ATTRIBUTIONS.md`** for upstream attributions and component-specific licenses before reusing or redistributing source imagery and geospatial files.
+
+## 📖 Documentation and paper
+
+The dataset package includes **`user-guide-final.md`**, the authoritative guide covering the benchmark contract, all 43 question types, runtime/evaluation behavior, and release setup. For executable inference-time code and additional project documentation, see the [GeoDisaster GitHub repository](https://github.com/VIMAGE-IITB/GeoDisaster).
+
+**Paper:** *GeoDisaster: Benchmarking Orchestrated Agents for Operational Disaster Geo-Intelligence* — **accepted at BMVC 2026**.
+
+**Authors:** Maram Hasan, Aman Verma, Savitra Roy, Hariseetharam Gunduboina, Daksh Jain, Muhammad Haris Khan, Subhasis Chaudhuri, and Biplab Banerjee.
+
+**[📄 Read PDF](https://arxiv.org/pdf/2606.17246)** · **[🔎 arXiv abstract](https://arxiv.org/abs/2606.17246)** · **[💻 Code](https://github.com/VIMAGE-IITB/GeoDisaster)**
+
+### Citation
+
+If you use GeoDisaster tasks, reference outputs, or trajectories, please cite the accompanying paper:
+
+```bibtex
+@misc{hasan2026geodisaster,
+  title         = {GeoDisaster: Benchmarking Orchestrated Agents for Operational Disaster Geo-Intelligence},
+  author        = {Maram Hasan and Aman Verma and Savitra Roy and Hariseetharam Gunduboina and Daksh Jain and Muhammad Haris Khan and Subhasis Chaudhuri and Biplab Banerjee},
+  year          = {2026},
+  eprint        = {2606.17246},
+  archivePrefix = {arXiv},
+  primaryClass  = {cs.CV},
+  url           = {https://arxiv.org/abs/2606.17246},
+  note          = {Accepted at BMVC 2026}
+}
+```
+
+