Datasets:

Modalities:
Text
Formats:
json
Languages:
English
ArXiv:
Libraries:
Datasets
pandas
License:
Daniel Paleka commited on
Commit
6b5ad6b
·
1 Parent(s): f98b5ee

Update README

Browse files
Files changed (1) hide show
  1. README.md +6 -6
README.md CHANGED
@@ -18,12 +18,12 @@ pretty_name: WildChat-2k-TypeTopic
18
  WildChat-2k-TypeTopic is a curated subset of single-message user prompts is constructed as follows:
19
 
20
  1. Filter out (using an LLM filter) prompts that:
21
- are not in English
22
- are not meaningful tasks (e.g., random character strings, “hello”)
23
- are incomplete (e.g., “Fix this code” with no code provided)
24
- are infeasible for text-only LLMs (e.g., “Describe a time when you worked in a team”, “an image of a cat”)
25
- are clearly part of multi-turn conversations (e.g., text-based game setups)
26
- are more than 800 characters long
27
 
28
  2. Deduplicate using `text-embedding-3-large` embeddings.
29
  3. Classify them into 16 task types and 25 topic categories; and subsample 2000 tasks to preserve representation of all types and categories;
 
18
  WildChat-2k-TypeTopic is a curated subset of single-message user prompts is constructed as follows:
19
 
20
  1. Filter out (using an LLM filter) prompts that:
21
+ * are not in English
22
+ * are not meaningful tasks (e.g., random character strings, “hello”)
23
+ * are incomplete (e.g., “Fix this code” with no code provided)
24
+ * are infeasible for text-only LLMs (e.g., “Describe a time when you worked in a team”, “an image of a cat”)
25
+ * are clearly part of multi-turn conversations (e.g., text-based game setups)
26
+ * are more than 800 characters long
27
 
28
  2. Deduplicate using `text-embedding-3-large` embeddings.
29
  3. Classify them into 16 task types and 25 topic categories; and subsample 2000 tasks to preserve representation of all types and categories;