Fangjun commited on
Commit
0cfedc2
Β·
verified Β·
1 Parent(s): 284ad31

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +75 -1
README.md CHANGED
@@ -9,4 +9,78 @@ tags:
9
  - language model evaluation
10
  size_categories:
11
  - 1K<n<10K
12
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9
  - language model evaluation
10
  size_categories:
11
  - 1K<n<10K
12
+ ---
13
+
14
+
15
+ ```markdown
16
+ # RoomSpace Dataset <img src="https://cdn-uploads.huggingface.co/production/uploads/61428f9ff0d914a5f606a82a/yMiDaPIIGOkmj3YrUBQ9E.jpeg" alt="RoomSpace" style="width: 80px; vertical-align: middle; margin-left: 10px;">
17
+
18
+ ## Description
19
+
20
+ The RoomSpace dataset, associated with the IJCAI-24 paper titled "Reframing Spatial Reasoning Evaluation in Language Models: A Real-World Simulation Benchmark for Qualitative Reasoning," aims to enhance the evaluation of spatial reasoning in language models. This benchmark encompasses a comprehensive range of qualitative spatial relationships, including topological, directional, and distance relations. These relationships are presented from various viewpoints, with differing levels of granularity and density of relational constraints to simulate real-world complexities. This approach promotes a more accurate assessment of language models' capabilities in spatial reasoning tasks.
21
+
22
+
23
+ ## Directory Structure
24
+
25
+ The repository is organized as follows:
26
+
27
+ ```
28
+ Data/
29
+ β”œβ”€β”€ SD-100/
30
+ β”œβ”€β”€ SD-1K/
31
+ └── SD-10K/
32
+ ```
33
+
34
+ - `SD-100`: Contains data for 100 room samples.
35
+ - `SD-1K`: Contains data for 1,000 room samples, with the first 100 rooms identical to those in `SD-100`.
36
+ - `SD-10K`: Contains data for 10,000 room samples, with the first 1,000 rooms identical to those in `SD-1K`.
37
+
38
+ Each `SD-*` subdirectory includes the following structure:
39
+
40
+ ```
41
+ SD-*/
42
+ β”œβ”€β”€ SD-*.json
43
+ β”œβ”€β”€ Image/
44
+ β”‚ β”œβ”€β”€ top-down/
45
+ β”‚ β”œβ”€β”€ robot_south/
46
+ β”‚ └── robot_south_rec/
47
+ β”œβ”€β”€ Text/
48
+ └── Logic/
49
+ ```
50
+
51
+ ### SD-*.json
52
+
53
+ This JSON file contains settings for each scene within the respective subdirectory. It details the configuration of the rooms, the objects placed, and any specific parameters or constraints relevant to the simulations in that dataset.
54
+
55
+ ### Image
56
+
57
+ This directory includes subdirectories for different views of the rooms:
58
+
59
+ - `top-down`: Top-down images of the rooms.
60
+ - `robot_south`: North-facing square images, assuming a robot stands at the south door.
61
+ - `robot_south_rec`: North-facing rectangular images.
62
+
63
+ Each image file is named `ID.png`, corresponding to the room's ID.
64
+
65
+ ### Text
66
+
67
+ This directory contains JSON files named `n*_m*_d*.json`, which include stories, questions, and answers derived from the images and a Constraint Satisfaction Problem (CSP) reasoner. The parameters `n`, `m`, and `d` represent the object number, constraint density, and domain size, respectively.
68
+
69
+ ### Logic
70
+
71
+ This directory includes JSON files under different spatial reasoning settings such as 'Layout', 'TPP', 'O2', 'O2+D2', 'O2+D3', 'Layout+O2+D2', and 'Layout+O2+D3'. These files contain logical facts and reasoning results:
72
+
73
+ - `n*_m*_d*.json`: Based on `n`, `m`, and `d` parameters as described.
74
+ - `answers_lengths_d*.json`: Records the answer lengths for each setting and each example for different `d` values.
75
+ - `times_fr_take_d*.json`: CPU time taken for each setting of the find relation question.
76
+ - `times_yn_take_d*.json`: CPU time taken for each setting of the yes/no question.
77
+ - `skip_id_d*.json`: Records the example IDs that do not meet the `m` requirements for the object numbers, which will be skipped.
78
+ - `solution_id_d*.json`: Records the IDs that have solutions (answer length > 0) under each spatial setting.
79
+
80
+ ## Citation
81
+
82
+
83
+ Fangjun Li, David Hogg, and Anthony Cohn. "Reframing Spatial Reasoning Evaluation in Language Models: A Real-World Simulation Benchmark for Qualitative Reasoning." IJCAI, 2024.
84
+
85
+ Fangjun Li, David Hogg, and Anthony Cohn (2024). RoomSpace. [Dataset]. https://doi.org/10.5518/1518
86
+ ```