Update README.md
Browse files
README.md
CHANGED
|
@@ -47,7 +47,7 @@ tags:
|
|
| 47 |
|
| 48 |
# ExtractBench Pro
|
| 49 |
|
| 50 |
-
**Quick links:** [\[🌐 Website\]](https://
|
| 51 |
|
| 52 |
ExtractBench Pro is the hard companion to [ExtractBench](https://huggingface.co/datasets/llamaindex/ExtractBench). The task is the same: given a full document and a user-written schema, return a schema-valid JSON object with the correct, exhaustive values and evidence for each one. The difference is what makes a value hard. In ExtractBench Pro, many values cannot be copied from the page. The reader has to read a curve or a drawing, apply the counting rules the issuer states once, or follow a convention set elsewhere in the document.
|
| 53 |
|
|
|
|
| 47 |
|
| 48 |
# ExtractBench Pro
|
| 49 |
|
| 50 |
+
**Quick links:** [\[🌐 Website\]](https://extractbenchpro.ai) [\[💻 Code\]](https://github.com/run-llama/ExtractBench) [\[📚 ExtractBench\]](https://huggingface.co/datasets/llamaindex/ExtractBench)
|
| 51 |
|
| 52 |
ExtractBench Pro is the hard companion to [ExtractBench](https://huggingface.co/datasets/llamaindex/ExtractBench). The task is the same: given a full document and a user-written schema, return a schema-valid JSON object with the correct, exhaustive values and evidence for each one. The difference is what makes a value hard. In ExtractBench Pro, many values cannot be copied from the page. The reader has to read a curve or a drawing, apply the counting rules the issuer states once, or follow a convention set elsewhere in the document.
|
| 53 |
|