LLMSQL: Upgrading WikiSQL for the LLM Era of Text-to-SQL

Compactor has been introduced as a training-free, query-agnostic key-value (KV) cache compression strategy for large language models (LLMs), utilizing approximate leverage scores to assess token importance. This method allows for a reduction of 20% in token retention while maintaining performance across various tasks, achieving a 68% reduction in KV memory burden on average.

Recent empirical studies have explored the capabilities of slow-thinking large language models (LLMs) like DeepSeek-R1 and ChatGPT-o1 in time series forecasting (TSF), proposing a new framework called TimeReasoner that treats TSF as a conditional reasoning task. This approach aims to enhance the models' ability to reason over temporal patterns, potentially improving forecasting accuracy even in zero-shot scenarios.

MobileFineTuner has been introduced as a unified open-source framework that enables end-to-end fine-tuning of large language models (LLMs) directly on commodity mobile phones, addressing the gap in existing methods that primarily rely on simulation or IoT devices.

Recent research indicates that large language models (LLMs) can enhance their reasoning capabilities through pure reinforcement learning (RL) focused on problem-solving, without the need for process reward models (PRMs). This finding challenges the traditional belief that PRMs are essential for developing reasoning skills in LLMs, as demonstrated by the DeepSeek-R1 model.