Add question type description to readme

README.md

@@ -60,39 +60,39 @@ One JSON object per line of `public.jsonl`:
 
 ## Question types
 
-| question_type | public | private |
-|---|---|---|
-| `contrastive-hard-negative/false-premise-after` | 39 | 9 |
-| `contrastive-hard-negative/false-premise-visual-attribute` | 47 | 9 |
-| `contrastive-hard-negative/transcript_visual` | 42 | 10 |
-| `contrastive-hard-negative/visual_audio` | 35 | 7 |
-| `distinct_count` | 2 | 1 |
-| `global-graph/co_presence` | 20 | 1 |
-| `global-graph/interaction_partner` | 17 | 7 |
-| `global-graph/most_frequent` | 11 | 6 |
-| `global-graph/never_co_present` | 21 | 1 |
-| `infinibench/causal_attribution_with_rules` | 1 | 0 |
-| `infinibench/collision_prediction` | 13 | 3 |
-| `infinibench/counting_cross_frame` | 12 | 2 |
-| `infinibench/lossy_compression` | 13 | 2 |
-| `infinibench/needle_in_the_haystack` | 1 | 2 |
-| `infinibench/performing_3d_to_2d_projections` | 12 | 4 |
-| `infinibench/persistent_entity_movement_tracking` | 13 | 2 |
-| `infinibench/persistent_entity_movement_tracking_counting_cross_frame` | 0 | 1 |
-| `infinibench/persistent_entity_movement_tracking_positional_memory` | 1 | 0 |
-| `infinibench/positional_memory` | 9 | 8 |
-| `infinibench/predicting_outcomes_of_constructions` | 5 | 0 |
-| `infinibench/size_comparisons` | 1 | 0 |
-| `k-hop/3_hop` | 18 | 5 |
-| `k-hop/4_hop` | 3 | 0 |
-| `k-hop/5_hop` | 3 | 2 |
-| `k-hop/7_hop` | 2 | 0 |
-| `longvid_pointer` | 0 | 1 |
-| `near-duplicate-moments` | 25 | 7 |
+| question_type | description |public | private |
+|---|---|---|---|
+| `contrastive-hard-negative/false-premise-after` | Question contains a false premise | 39 | 9 |
+| `contrastive-hard-negative/false-premise-visual-attribute` | Question contains a false premise | 47 | 9 |
+| `contrastive-hard-negative/transcript_visual` | Question contains a false premise |42 | 10 |
+| `contrastive-hard-negative/visual_audio` | Question contains a false premise| 35 | 7 |
+| `distinct_count` | Counting distinct objects |2 | 1 |
+| `global-graph/co_presence` | Identify entites that often co-occur across the whole video |20 | 1 |
+| `global-graph/interaction_partner`| Identify entites that often interact across the whole video | 17 | 7 |
+| `global-graph/most_frequent` | Identify events that happen more frequently then others across the whole video | 11 | 6 |
+| `global-graph/never_co_present` | Identify entites that never co-occur across the whole video | 21 | 1 |
+| `infinibench/causal_attribution_with_rules` | Describe outcome of a state based on a given rule | 1 | 0 |
+| `infinibench/collision_prediction` | Extrapolate object trajectory in zero-acceleration settings | 13 | 3 |
+| `infinibench/counting_cross_frame` | Counting objects across multiple frames | 12 | 2 |
+| `infinibench/lossy_compression` | Identifying a small relevant subset of entities in a cluttered scene | 13 | 2 |
+| `infinibench/needle_in_the_haystack` | Identify spurious visual detail | 1 | 2 |
+| `infinibench/performing_3d_to_2d_projections` | Understand 3D space from different perspectives | 12 | 4 |
+| `infinibench/persistent_entity_movement_tracking` | Follow one or more entites while the change across frames | 13 | 2 |
+| `infinibench/persistent_entity_movement_tracking_counting_cross_frame`| Follow one or more entites while the change across frames and count| 0 | 1 |
+| `infinibench/persistent_entity_movement_tracking_positional_memory` | Recall position of object and follow moving entity | 1 | 0 |
+| `infinibench/positional_memory` | Identify positions of memory cards when revealing them one-by-one | 9 | 8 |
+| `infinibench/predicting_outcomes_of_constructions`| Predict visual appearence of an object after rotation/transformation | 5 | 0 |
+| `infinibench/size_comparisons`| Compare relative sizes of objects | 1 | 0 |
+| `k-hop/3_hop` | Follow chain of audio-visual cues through the video, takes 3 reasoning steps | 18 | 5 |
+| `k-hop/4_hop` | Follow chain of audio-visual cues through the video, takes 4 reasoning steps | 3 | 0 |
+| `k-hop/5_hop` | Follow chain of audio-visual cues through the video, takes 5 reasoning steps | 3 | 2 |
+| `k-hop/7_hop` | Follow chain of audio-visual cues through the video, takes 7 reasoning steps | 2 | 0 |
+| `longvid_pointer` | Answer Question about long video | 0 | 1 |
+| `near-duplicate-moments` | Anwer question about one particular moment in a video that contains another similar moment | 25 | 7 |
 | `reidentification` | 2 | 1 |
-| `same_or_different` | 29 | 5 |
-| `same_or_different/graphic` | 2 | 1 |
-| `same_or_different/scene` | 1 | 0 |
+| `same_or_different` | Determine if two entities are the same | 29 | 5 |
+| `same_or_different/graphic` | Determine if two entities are the same| 2 | 1 |
+| `same_or_different/scene`| Determine if two entities are the same | 1 | 0 |
 | `sequential/first_seen` | 39 | 14 |
 | **total** | **439** | **111** |