NJUNLP

Code and data for RP-OPSD main experiments

arXivLiveBench: A Live Benchmark for Long-Context Modelling

Recognize Your Orchestrator: An Entropy Dynamics Perspective for LLM Multi-Agent Systems

The code and datasets of our paper "Friend or Foe: How LLMs' Safety Mind Gets Fooled by Intent Shift Attack".

CogGen: A Cognitively Inspired Recursive Framework for Deep Research Report Generation

This is the official implementation of the paper PEGRL: Improving Machine Translation by Post-Editing Guided Reinforcement Learning.

A novel Group Relative Reward Model (GRRM) framework enhances machine translation quality and reasoning capabilities by improving intra-group ranking through comparative analysis rather than isolated metric evaluation.