Upload README.md with huggingface_hub

README.md

@@ -6,57 +6,38 @@ tags:
 - curriculum-learning
 - difficulty-estimation
 - rule-based
-dataset_info:
-  features:
-  - name: image
-    dtype: image
-  - name: user
-    dtype: string
-  - name: assistant
-    dtype: string
-  - name: difficulty_tier
-    dtype: string
-  splits:
-  - name: train
-    num_bytes: 7194085718.375
-    num_examples: 199293
-  download_size: 121124922
-  dataset_size: 7194085718.375
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
 ---
 
-# Cleaned-PlotQA v2 difficulty tiers (rule-based)
+# Cleaned-PlotQA v2 with difficulty tiers (balanced, rule-based)
 
-This repository augments jrc/cleaned-plotqa-v2 with a single additional column, difficulty_tier ∈ {easy, medium, hard}, tailored to the numeric and visual reasoning characteristics of PlotQA-style scientific plots [1][2][3].  
+This repository augments jrc/cleaned-plotqa-v2 with one additional column, difficulty_tier ∈ {easy, medium, hard}. The classifier uses strong PlotQA-oriented rules emphasizing numeric reasoning, extremum/trend/slope semantics, visual grounding (axes/legend/lines/bars), units/notation, and multi-entity disambiguation. Tiers are balanced using quantiles computed over the full dataset for robust stage sizing in curriculum learning.
 
 ## Tier counts
 
-- easy: 199293  
-- medium: 0  
-- hard: 0  
-- total labeled: 199293  
+- easy: 199293
+- medium: 0
+- hard: 0
+- total labeled: 199293
 
 ## Criteria summary
 
-This release adds a single column difficulty_tier ∈ {easy, medium, hard} using a deterministic, PlotQA-oriented scoring function:
-- Numeric reasoning: detection of arithmetic/aggregation (sum, difference, ratio, percentage, average), multi-operator stacking, and comparison/threshold cues.
-- Extremum/trend: explicit emphasis on max/min/peak/valley and slope/rate-of-change/derivative; any “line/curve + extremum” is never classified as easy.
+This release adds a single column difficulty_tier ∈ {easy, medium, hard} using a deterministic, PlotQA-oriented scoring function with balanced thresholds:
+
+Scoring highlights:
+- Numeric reasoning: detection of arithmetic/aggregation (sum, difference, ratio, percentage, average), stacked operations, and comparison/threshold cues.
+- Extremum/trend/slope: emphasis on max/min/peak/valley and slope/rate-of-change; any “line/curve + extremum” is never classified as easy.
 - Visual grounding: references to axes, ticks, legend, lines/bars/curves, and positional terms (left/right/top/bottom/adjacent) raise difficulty.
 - Units and formats: recognition of %, scientific notation, unit tokens (k/M/B, °C, km, kg, Hz, etc.), ranges (10–20), and uncertainty (±, std dev, variance).
 - Series disambiguation: color terms and multi-entity words (both/all/each/every/together/combined/grouped/stacked/multi) increase difficulty.
-Thresholds are tuned higher than for ChartQA to reflect stronger numeric complexity commonly observed in scientific plot QA tasks.
-
-## Notes
 
-- Only one new column is introduced; no changes to the original fields of the cleaned dataset.  
-- The rules prioritize numeric reasoning, extremum/trend detection, units, and multi-entity aggregation for realistic difficulty separation in PlotQA-style QA [2][3].  
+Balancing:
+- Tiers are assigned by data-driven quantiles over the entire dataset:
+  - easy: score ≤ 0.0000 (≈ bottom 40%)
+  - medium: 0.0000 < score ≤ 0.0000 (≈ middle 40%)
+  - hard: score > 0.0000 (≈ top 20%)
+This ensures a balanced distribution even if raw scores are skewed.
 
-## References
+## Notes
 
-[1] jrc/cleaned-plotqa-v2 (compact cleaned release for convenient QA experimentation).  
-[2] PlotQA: Reasoning over Scientific Plots (benchmark introducing large-scale, open-vocabulary numeric QA on plots).  
-[3] PlotQA details and motivation for numeric reasoning challenges in open-vocabulary answer settings.
+- Only one new column is introduced; all original fields remain unchanged.
+- The balancing by quantiles is designed to produce sensible stage sizes across easy/medium/hard and can be retuned (e.g., 30/40/30) by adjusting the quantile cut points.