Update README.md

README.md

@@ -43,12 +43,12 @@ configs:
 
 ## Dataset Summary
 
-This dataset is a **filtered subset of MuirBench**, containing only samples with **exactly 2 or 3 input images per question**.
+This dataset is a **filtered subset of MuirBench**, containing only samples with **exactly 2 input images per question**.
 
 The goal of this subset is to support controlled experiments on **multi-image visual reasoning**, where the number of images is fixed and small, enabling clearer analysis of compositional and relational reasoning in Vision-Language Models (VLMs).
 
 - Original dataset: **MuirBench**
-- Filter criterion: `len(image_list) ∈ {2, 3}`
+- Filter criterion: `len(image_list) ∈ {2}`
 - Modality: Image + Text
 - Task type: Multi-image visual reasoning / VQA-style QA
 
@@ -62,7 +62,7 @@ Each sample contains (a subset of the original fields):
 |----------------|-------------------|-------------|
 | `question`     | string            | The input question |
 | `image_list`   | list of images    | A list of 2 or 3 images (stored as bytes) |
-| `count`        | int               | Number of images (`2` or `3`) |
+| `count`        | int               | Number of images (`2`) |
 | `choices`*     | list (optional)   | Multiple-choice options (if applicable) |
 | `answer`*      | string/int        | Ground-truth answer |
 | `task`*        | string            | Task category |
@@ -92,7 +92,7 @@ This dataset was created by:
 
 1. Loading the original MuirBench dataset
 2. Counting the number of images per sample (`len(image_list)`)
-3. Filtering samples where the image count is exactly 2 or 3
+3. Filtering samples where the image count is exactly 2 
 4. Releasing the filtered subset as a standalone dataset for research convenience
 
 No annotations were modified.