Datasets:

Modalities:
Text
Languages:
English
Size:
< 1K
ArXiv:
License:
ronch99 commited on
Commit
b9cefee
·
verified ·
1 Parent(s): 5c930ba

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +6 -0
README.md CHANGED
@@ -47,6 +47,12 @@ dataset_info:
47
 
48
  ## ScienceAgentBench
49
 
 
 
 
 
 
 
50
  The advancements of language language models (LLMs) have piqued growing interest in developing LLM-based language agents to automate scientific discovery end-to-end, which has sparked both excitement and skepticism about their true capabilities.
51
  In this work, we call for rigorous assessment of agents on individual tasks in a scientific workflow before making bold claims on end-to-end automation.
52
  To this end, we present ScienceAgentBench, a new benchmark for evaluating language agents for data-driven scientific discovery:
 
47
 
48
  ## ScienceAgentBench
49
 
50
+ ### Update 04/30/2026: To mitigate false negatives in evaluation, we have released a verified version of ScienceAgentBench. Please load our benchmark using the following code going forward:
51
+ ```py
52
+ from datasets import load_dataset
53
+ ds = load_dataset("osunlp/ScienceAgentBench", split="verified")
54
+ ```
55
+
56
  The advancements of language language models (LLMs) have piqued growing interest in developing LLM-based language agents to automate scientific discovery end-to-end, which has sparked both excitement and skepticism about their true capabilities.
57
  In this work, we call for rigorous assessment of agents on individual tasks in a scientific workflow before making bold claims on end-to-end automation.
58
  To this end, we present ScienceAgentBench, a new benchmark for evaluating language agents for data-driven scientific discovery: