ROUGE-L

Text overlap measure based on the longest common subsequence between candidate and reference.

Findings

• IC-005GPT-4o underperforms AHA and other VLMs in detecting and reasoning about robotic manipulation failures across multiple datasets. [eval]
• IC-304Instruction-tuned LMs become more vulnerable to prompt-injected data extraction as model size increases from 7B to 70B [eval]
• IC-306Instruction tuning increases the ROUGE score of prompt-injected data extraction by 65.76 on average compared to base models [eval]
• IC-730CLIPCap and BLIP-2 produce degraded alt-text on Twitter social media images, with BLEU@4 of 0.372 and 0.111 respectively [eval]
• IC-812InstructGPT (text-davinci-003) reduces content diversity in co-written essays while GPT-3 (davinci) does not, and the effect is attributable to the model's own less diverse text contributions [eval]