Restructure into data/ subfolder, update README with SAHF methodology and dataset configs

- Move data files into data/ subfolder
- Add configs block to README YAML to register votes and market_summary as separate viewer tabs
- Update dataset card with stake-assured human feedback framing, voting power stats, limitations section, and whitepaper link

DATA_DICTIONARY.md

@@ -19,7 +19,7 @@ Row-level anonymized crowd feedback votes.
 | `vote` | `up_vote` or `down_vote` |
 | `up_vote` | Boolean vote direction |
 | `votes` | Numeric vote count from source system |
-| `voting_power` | Voting weight from source system |
+| `voting_power` | Weight attached to the feedback vote |
 | `created_at` | Original creation timestamp |
 | `updated_at` | Original update timestamp |
 | `deleted` | Whether the source row was deleted |
@@ -48,6 +48,19 @@ One row per geopolitical question/market, aggregated from `data/train.jsonl`.
 | `last_seen_at` | Latest feedback timestamp |
 | `split` | Dataset split |
 
+## Recommended training setup
+
+For a simple supervised setup:
+
+- Input: `question` + `feedback`
+- Label: `up_vote` or `vote`
+- Sample weight: `voting_power`
+
+For market-level ranking or triage:
+
+- Input: `question`
+- Targets/features: `up_vote_share`, `num_feedback_votes`, `total_voting_power`, `top_feedback_tags`
+
 ## Privacy changes
 
 The source CSV contained `voter_id` and `private_subnet_id`. Those fields are intentionally excluded from this public release.

README.md

@@ -1,3 +1,38 @@
+---
+license: cc-by-4.0
+language:
+- en
+tags:
+- prediction-markets
+- crowdsourcing
+- geopolitics
+- forecasting
+- reasoning
+- ai-evaluation
+- agents
+- uncertainty
+- calibration
+- human-feedback
+- stake-assured
+- rlhf
+task_categories:
+- text-classification
+- question-answering
+- text-generation
+pretty_name: MarketCrowd Geopolitics
+size_categories:
+- n<1K
+configs:
+- config_name: votes
+  data_files:
+    - split: train
+      path: data/train.jsonl
+- config_name: market_summary
+  data_files:
+    - split: train
+      path: data/market_summary.jsonl
+---
+
 # MarketCrowd Geopolitics
 
 **The first open dataset produced via stake-assured human feedback (SAHF)** — preference signals crowdsourced through capital-at-risk voting on geopolitical AI reasoning.
@@ -6,11 +41,11 @@
 
 ## Overview
 
-MarketCrowd Geopolitics contains anonymized crowd feedback votes and market-level summaries derived from a geopolitical prediction-market workflow on the [Reppo protocol](https://reppo.xyz). 
+MarketCrowd Geopolitics contains anonymized crowd feedback votes and market-level summaries derived from a geopolitical prediction-market workflow on the [Reppo protocol](https://reppo.xyz).
 
 Unlike standard annotation datasets where labelers are paid per task, every signal in this dataset was produced by voters who locked $REPPO tokens as economic collateral — meaning their judgments carry capital at risk, not just attention.
 
-This is an **initial seed release** covering ~6 weeks of activity across ~20 geopolitical markets (March–April 2026). The underlying dataset updates continuously as new epochs settle on the Reppo protocol.
+This is an **initial seed release** covering ~6 weeks of activity across 51 geopolitical markets (March–April 2026). The underlying dataset updates continuously as new epochs settle on the Reppo protocol.
 
 ---
 
@@ -26,85 +61,82 @@ This dataset was generated through a [Reppo Datanet](https://reppo.xyz/reppo-whi
 
 This mechanism is described in full in the [Reppo whitepaper](https://reppo.xyz/reppo-whitepaper.pdf).
 
-### Why stake-assured feedback is different from standard annotation
+### Why stake-assured feedback differs from standard annotation
 
 | Property | Paid annotation (e.g. Scale AI) | Reppo SAHF |
 |---|---|---|
 | Voter incentive | Payment per task | Emissions + locked capital at risk |
-| Sybil resistance | Identity verification | Capital requirement (can't split stake to gain power) |
+| Sybil resistance | Identity verification | Capital requirement (splitting stake doesn't increase power) |
 | Disagreement handling | Averaged out | Explicit signal via net-vote mechanism |
 | Quality metric | Inter-annotator agreement | Economic Value of Feedback (EVOF) |
 | Data staleness | Static snapshot | Continuously updated every 48 hours |
 
----
-
-## What Is Included
+### Voting power concentration
 
-This release includes four data files:
+The `voting_power` field is derived from locked $REPPO × lock duration (non-linear). In this release, voting power ranges from 2,471 to 3,013,340 with a median of 64,110. The top 3 wallets account for ~13% of total voting power — relatively distributed for a token-weighted system. The Reppo protocol further mitigates concentration via square-root dampening in the EVOF metric.
 
-| File | Description |
-|---|---|
-| `train.jsonl` | Row-level anonymized crowd feedback votes |
-| `train.csv` | CSV version of the row-level data |
-| `market_summary.jsonl` | Aggregated summaries by geopolitical question/market |
-| `market_summary.csv` | CSV version of the market-level summary data |
+---
 
-Additional documentation:
+## What Is Included
 
-| File | Description |
-|---|---|
-| `DATA_DICTIONARY.md` | Explains each field in the dataset |
-| `dataset_schema.json` | Machine-readable schema |
+| File | Rows | Description |
+|---|---:|---|
+| `data/train.jsonl` | 160 | Row-level anonymized weighted crowd feedback votes |
+| `data/train.csv` | 160 | CSV version of the row-level data |
+| `data/market_summary.jsonl` | 51 | Aggregated summaries by geopolitical question/market |
+| `data/market_summary.csv` | 51 | CSV version of the market-level summary data |
+| `DATA_DICTIONARY.md` | — | Field-level documentation |
+| `dataset_schema.json` | — | Machine-readable schema |
 
 ---
 
 ## Dataset Structure
 
-### Row-level file: `train.jsonl`
+### `data/train.jsonl` — row-level votes
 
 Each row is an anonymized feedback vote attached to a geopolitical prediction-market question.
 
-| Field | Description |
-|---|---|
-| `id` | Stable public row ID |
-| `source_record_id` | Original feedback record ID |
-| `record_type` | `crowd_feedback_vote` |
-| `question` | Geopolitical market/question title |
-| `category` | Topic category |
-| `market_id` | Original market/Pod ID |
-| `epoch` | 48-hour epoch in which the vote was cast (higher = more recent) |
-| `feedback` | Qualitative feedback label and/or freeform comment. Format: `"Label1, Label2 - optional comment"`. Note: this field mixes structured tags with open-ended text. |
-| `vote` | `up_vote` or `down_vote` |
-| `up_vote` | Boolean vote direction |
-| `votes` | Raw on-chain vote count from source system. Note: this is distinct from `voting_power` and will be 0 for many rows where voting power was allocated without a discrete on-chain vote count being recorded. |
-| `voting_power` | veREPPO weight applied to this vote — derived from $REPPO locked × lock duration (non-linear). This is the economically meaningful signal. |
-| `created_at` | Feedback creation timestamp |
-| `updated_at` | Feedback update timestamp |
-| `deleted` | Source deletion flag |
-| `version` | Source version field |
-| `split` | Dataset split |
-
-### Market-level file: `market_summary.jsonl`
-
-Each row summarizes all available feedback for one geopolitical question/market.
-
-| Field | Description |
-|---|---|
-| `id` | Stable public summary ID |
-| `record_type` | `market_feedback_summary` |
-| `question` | Geopolitical market/question title |
-| `category` | Topic category |
-| `market_id` | Original market/Pod ID |
-| `num_feedback_votes` | Number of feedback votes |
-| `up_votes` | Count of positive votes |
-| `down_votes` | Count of negative votes |
-| `up_vote_share` | Share of positive votes |
-| `total_voting_power` | Sum of voting power across all votes |
-| `median_voting_power` | Median voting power per vote |
-| `top_feedback_tags` | Most common feedback labels/comments |
-| `first_seen_at` | Earliest feedback timestamp |
-| `last_seen_at` | Latest feedback timestamp |
-| `split` | Dataset split |
+| Field | Type | Description |
+|---|---|---|
+| `id` | string | Stable public row ID |
+| `source_record_id` | string | Original feedback record ID |
+| `record_type` | string | Always `crowd_feedback_vote` |
+| `question` | string | Geopolitical market/question title |
+| `category` | string | Always `geopolitics` in this release |
+| `market_id` | string | Original market/Pod ID |
+| `epoch` | int | 48-hour epoch in which the vote was cast (higher = more recent) |
+| `feedback` | string | Qualitative label and/or freeform comment. Format: `"Label1, Label2 - optional comment"` |
+| `vote` | string | `up_vote` or `down_vote` |
+| `up_vote` | bool | Boolean vote direction |
+| `votes` | int | Raw on-chain vote count. Distinct from `voting_power` — will be 0 for many rows |
+| `voting_power` | int | veREPPO weight — the economically meaningful signal. Derived from $REPPO locked × lock duration |
+| `created_at` | string | Feedback creation timestamp |
+| `updated_at` | string | Feedback update timestamp |
+| `deleted` | bool | Source deletion flag |
+| `version` | int | Source version field |
+| `split` | string | Dataset split |
+
+### `data/market_summary.jsonl` — market-level aggregates
+
+One row per geopolitical question, summarizing all votes for that market.
+
+| Field | Type | Description |
+|---|---|---|
+| `id` | string | Stable public summary ID |
+| `record_type` | string | Always `market_feedback_summary` |
+| `question` | string | Geopolitical market/question title |
+| `category` | string | Topic category |
+| `market_id` | string | Original market/Pod ID |
+| `num_feedback_votes` | int | Total number of feedback votes |
+| `up_votes` | int | Count of positive votes |
+| `down_votes` | int | Count of negative votes |
+| `up_vote_share` | float | Fraction of votes that were positive |
+| `total_voting_power` | int | Sum of voting power across all votes |
+| `median_voting_power` | float | Median voting power per vote |
+| `top_feedback_tags` | list | Most common feedback labels |
+| `first_seen_at` | string | Earliest feedback timestamp |
+| `last_seen_at` | string | Latest feedback timestamp |
+| `split` | string | Dataset split |
 
 ---
 
@@ -113,12 +145,13 @@ Each row summarizes all available feedback for one geopolitical question/market.
 ```json
 {
   "id": "mcg_vote_000001",
+  "source_record_id": "cmndwvyey0001kz04mn8e5fmp",
   "record_type": "crowd_feedback_vote",
   "question": "Hormuz Updates",
   "category": "geopolitics",
   "market_id": "cmn9plf2k0001kz040cuuizdq",
   "epoch": 64,
-  "feedback": "High quality - ",
+  "feedback": "High quality -",
   "vote": "up_vote",
   "up_vote": true,
   "votes": 0,
@@ -135,61 +168,61 @@ Each row summarizes all available feedback for one geopolitical question/market.
 
 ## Suggested Uses
 
-This dataset is best used as a **feedback-signal and question-quality dataset**. It should not be treated as a standalone source of resolved geopolitical ground truth.
+This dataset is best treated as a **feedback-signal and question-quality dataset**. It should not be used as a source of resolved geopolitical ground truth.
 
-Suitable applications include:
+**Recommended training setup:**
 
-- Text classification of feedback labels
-- Question-quality scoring and market triage
-- Weak supervision from capital-weighted crowd votes
-- Market/question ranking by crowd-perceived usefulness
-- Crowd-signal summarization
-- Geopolitical forecasting prompt evaluation
-- Research into prediction-market-based and stake-assured data generation
-- Evaluation of agents that reason about market questions and crowd feedback
+| Component | Field |
+|---|---|
+| Input | `question` + `feedback` |
+| Label | `up_vote` or `vote` |
+| Sample weight | `voting_power` |
 
-### Agent Training Use Cases
+**Agent training use cases:**
 
-| Agent task | Input | Training or evaluation signal |
+| Agent task | Input | Signal |
 |---|---|---|
-| Question quality scoring | Geopolitical market question | Up-vote share, voting power, feedback labels |
-| Market triage | Market question + feedback | Aggregated market summary |
-| Feedback classification | Feedback text | Crowd feedback category or vote direction |
+| Question quality scoring | Market question | `up_vote_share`, `voting_power`, feedback labels |
+| Market triage | Question + feedback | Aggregated market summary |
+| Feedback classification | Feedback text | Vote direction or feedback category |
 | Signal summarization | Question + row-level votes | Market-level summary |
-| Forecasting workflow support | Proposed market/question | Crowd-perceived usefulness and clarity |
+| Forecasting workflow support | Proposed market | Crowd-perceived usefulness and clarity |
+
+**Other applications:** text classification, weak supervision, market ranking, crowd-signal summarization, geopolitical forecasting prompt evaluation, prediction-market-based data research.
 
 ---
 
 ## Limitations
 
-- **Size**: This is a seed release of 160 rows across ~20 geopolitical markets from a single 6-week window (March–April 2026). It is not yet suitable as a standalone benchmark.
-- **Voting power concentration**: As with any token-weighted system, a small number of high-veREPPO wallets contribute a disproportionate share of voting power. The `voting_power` field reflects this directly. The Reppo protocol addresses this via square-root dampening in the EVOF metric, but downstream users should be aware of concentration when weighting signals.
-- **Feedback field structure**: The `feedback` field mixes structured tags (e.g. "High quality") with freeform comments. Preprocessing is recommended before using this field for classification tasks.
-- **Single domain**: All records are `category: geopolitics` from a single Datanet. Cross-domain generalization should not be assumed.
-- **No resolved outcomes**: This dataset captures crowd quality judgments, not forecasting accuracy or event resolution. It is a preference/feedback dataset, not a ground-truth geopolitical dataset.
+- **Size**: 160 votes across 51 markets from a 6-week window. This is a seed release — not yet suitable as a standalone benchmark.
+- **Single domain**: All records are `category: geopolitics` from one Datanet. Cross-domain generalization should not be assumed.
+- **Feedback field**: Mixes structured tags with freeform comments. Preprocessing recommended before classification tasks.
+- **No resolved outcomes**: This is a preference/feedback dataset, not a ground-truth geopolitical dataset. Event resolution is not included.
+- **`votes` vs `voting_power`**: The `votes` field is 0 for most rows — it reflects a raw on-chain count that was not recorded for all votes. Use `voting_power` as the primary signal.
 
 ---
 
-## About Reppo
+## Privacy
 
-Reppo is a protocol for tokenized, continuously curated data production for reinforcement learning. Datanets are on-chain data markets where publishers contribute content and veREPPO holders provide stake-assured quality assessments. The resulting datasets update every 48 hours and are available for subscription via [repo.exchange](https://repo.exchange).
+The source data contained direct voter and subnet identifiers. These were removed before this public release:
 
-- Website: [reppo.xyz](https://reppo.xyz)
-- Whitepaper: [reppo.xyz/reppo-whitepaper.pdf](https://reppo.xyz/reppo-whitepaper.pdf)
+- `voter_id` — removed
+- `private_subnet_id` — removed
 
 ---
 
-## License
+## About Reppo
+
+Reppo is a protocol for tokenized, continuously curated data production for reinforcement learning. Datanets are on-chain data markets where publishers contribute content and veREPPO holders provide stake-assured quality assessments. Datasets update every 48 hours and are available for subscription via [repo.exchange](https://repo.exchange).
 
-[CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)
+- Website: [reppo.xyz](https://reppo.xyz)
+- Whitepaper: [reppo.xyz/reppo-whitepaper.pdf](https://reppo.xyz/reppo-whitepaper.pdf)
 
 ---
 
 ## Citation
 
-If you use this dataset, please cite:
-
-```
+```bibtex
 @dataset{reppo_marketcrowd_geopolitics_2026,
   author    = {Reppo Labs},
   title     = {MarketCrowd Geopolitics: Stake-Assured Human Feedback on Geopolitical AI Reasoning},
@@ -197,4 +230,10 @@ If you use this dataset, please cite:
   publisher = {Hugging Face},
   url       = {https://huggingface.co/datasets/Reppo-labs/marketcrowd-geopolitics}
 }
-```
+```
+
+---
+
+## License
+
+[CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) — attribution to Reppo Labs and the MarketCrowd Geopolitics dataset required.

dataset_schema.json

@@ -1,52 +1,23 @@
 {
   "dataset_name": "MarketCrowd Geopolitics",
-  "version": "0.1.0",
-  "source_file": "geopolitics-dump (2).csv",
-  "privacy_note": "voter_id and private_subnet_id were removed from the public export.",
-  "tables": {
-    "data/train.jsonl": {
-      "description": "Anonymized row-level crowd feedback votes on geopolitics markets/questions.",
-      "num_rows": 160,
-      "fields": [
-        "id",
-        "source_record_id",
-        "record_type",
-        "question",
-        "category",
-        "market_id",
-        "epoch",
-        "feedback",
-        "vote",
-        "up_vote",
-        "votes",
-        "voting_power",
-        "created_at",
-        "updated_at",
-        "deleted",
-        "version",
-        "split"
-      ]
-    },
-    "data/market_summary.jsonl": {
-      "description": "Aggregated market/question-level summaries computed from the row-level feedback votes.",
-      "num_rows": 51,
-      "fields": [
-        "id",
-        "record_type",
-        "question",
-        "category",
-        "market_id",
-        "num_feedback_votes",
-        "up_votes",
-        "down_votes",
-        "up_vote_share",
-        "total_voting_power",
-        "median_voting_power",
-        "top_feedback_tags",
-        "first_seen_at",
-        "last_seen_at",
-        "split"
-      ]
-    }
+  "version": "1.0.0",
+  "num_row_level_records": 160,
+  "num_market_summary_records": 51,
+  "license": "cc-by-4.0",
+  "privacy_note": "voter_id and private_subnet_id were removed from the public release.",
+  "recommended_training_setup": {
+    "input": "question + feedback",
+    "label": "up_vote or vote",
+    "sample_weight": "voting_power",
+    "market_level_weight_features": [
+      "total_voting_power",
+      "median_voting_power"
+    ]
+  },
+  "files": {
+    "data/train.jsonl": "Row-level anonymized weighted crowd feedback votes.",
+    "data/market_summary.jsonl": "Market/question-level summaries aggregated from train.jsonl.",
+    "data/train.csv": "CSV copy of row-level data.",
+    "data/market_summary.csv": "CSV copy of market-level summaries."
   }
 }