AI RESEARCH
On Stable Long-Form Generation: Benchmarking and Mitigating Length Volatility
arXiv CS.CL
•
ArXi:2605.01357v1 Announce Type: new Large Language Models (LLMs) excel at long-context understanding but exhibit significant limitations in long-form generation. Existing studies primarily focus on single-generation quality, generally overlooking the volatility of the output. This volatility not only leads to significant computational costs but also severely impacts the models' reliable application. To address this gap, our work unfolds in three stages: benchmarking, probing, and mitigation.