Qwen2.5 Technical Report

In this report, we introduce Qwen2.5, a comprehensive series of large language models (LLMs) designed to meet diverse needs. Compared to previous iterations, Qwen 2.5 has been significantly improved during both the pre-training and post-training stages. In terms of pre-training, we have scaled the high-quality pre-training datasets from the previous 7 trillion tokens to 18 trillion tokens. This provides a strong foundation for common sense, expert knowledge, and reasoning capabilities. In terms of post-training, we implement intricate supervised finetuning with over 1 million samples, as well as multistage reinforcement learning. Post-training techniques enhance human preference, and notably improve long text generation, structural data analysis, and instruction following. To handle diverse and varied use cases effectively, we present Qwen2.5 LLM series in rich sizes. Open-weight offerings include base and instruction-tuned models, with quantized versions available. In addition, for hosted solutions, the proprietary models currently include two mixture-of-experts (MoE) variants: Qwen2.5-Turbo and Qwen2.5-Plus, both available from Alibaba Cloud Model Studio. Qwen2.5 has demonstrated top-tier performance on a wide range of benchmarks evaluating language understanding, reasoning, mathematics, coding, human preference alignment, etc. Specifically, the open-weight flagship Qwen2.5-72B-Instruct outperforms a number of open and proprietary models and demonstrates competitive performance to the state-of-the-art open-weight model, Llama-3-405B-Instruct, which is around 5 times larger. Qwen2.5-Turbo and Qwen2.5-Plus offer superior cost-effectiveness while performing competitively against GPT-4o-mini and GPT-4o respectively. Additionally, as the foundation, Qwen2.5 models have been instrumental in training specialized models such as Qwen2.5-Math, Qwen2.5-Coder, QwQ, and multimodal models.

Inference-Time Decontamination: Reusing Leaked Benchmarks for Large Language Model Evaluation

The training process of large language models (LLMs) often involves varying degrees of test data contamination. Although current LLMs are achieving increasingly better performance on various benchmarks, their performance in practical applications does not always match their benchmark results. Leakage of benchmarks can prevent the accurate assessment of LLMs' true performance. However, constructing new benchmarks is costly, labor-intensive and still carries the risk of leakage. Therefore, in this paper, we ask the question, Can we reuse these leaked benchmarks for LLM evaluation? We propose Inference-Time Decontamination (ITD) to address this issue by detecting and rewriting leaked samples without altering their difficulties. ITD can mitigate performance inflation caused by memorizing leaked benchmarks. Our proof-of-concept experiments demonstrate that ITD reduces inflated accuracy by 22.9% on GSM8K and 19.0% on MMLU. On MMLU, using Inference-time Decontamination can lead to a decrease in the results of Phi3 and Mistral by 6.7% and 3.6% respectively. We hope that ITD can provide more truthful evaluation results for large language models.

Unified Active Retrieval for Retrieval Augmented Generation

In Retrieval-Augmented Generation (RAG), retrieval is not always helpful and applying it to every instruction is sub-optimal. Therefore, determining whether to retrieve is crucial for RAG, which is usually referred to as Active Retrieval. However, existing active retrieval methods face two challenges: 1. They usually rely on a single criterion, which struggles with handling various types of instructions. 2. They depend on specialized and highly differentiated procedures, and thus combining them makes the RAG system more complicated and leads to higher response latency. To address these challenges, we propose Unified Active Retrieval (UAR). UAR contains four orthogonal criteria and casts them into plug-and-play classification tasks, which achieves multifaceted retrieval timing judgements with negligible extra inference cost. We further introduce the Unified Active Retrieval Criteria (UAR-Criteria), designed to process diverse active retrieval scenarios through a standardized procedure. Experiments on four representative types of user instructions show that UAR significantly outperforms existing work on the retrieval timing judgement and the performance of downstream tasks, which shows the effectiveness of UAR and its helpfulness to downstream tasks.

A webcam-based machine learning approach for three-dimensional range of motion evaluation

Background. Joint range of motion (ROM) is an important quantitative measure for physical therapy. Commonly relying on a goniometer, accurate and reliable ROM measurement requires extensive training and practice. This, in turn, imposes a significant barrier for those who have limited in-person access to healthcare. Objective. The current study presents and evaluates an alternative machine learning-based ROM evaluation method that could be remotely accessed via a webcam. Methods. To evaluate its reliability, the ROM measurements for a diverse set of joints (neck, spine, and upper and lower extremities) derived using this method were compared to those obtained from a marker-based optical motion capture system. Results. Data collected from 25 healthy adults demonstrated that the webcam solution exhibited high test-retest reliability, with substantial to almost perfect intraclass correlation coefficients for most joints. Compared with the marker-based system, the webcam-based system demonstrated substantial to almost perfect inter-rater reliability for some joints, and lower inter-rater reliability for other joints (e.g., shoulder flexion and elbow flexion), which could be attributed to the reduced sensitivity to joint locations at the apex of the movement. Conclusions. The proposed webcam-based method exhibited high test-retest and inter-rater reliability, making it a versatile alternative for existing ROM evaluation methods in clinical practice and the tele-implementation of physical therapy and rehabilitation.

Multi-Task Pre-Training of Modular Prompt for Few-Shot Learning

Prompt tuning is a parameter-efficient approach to adapting pre-trained language models to downstream tasks. Although prompt tuning has been shown to match the performance of full model tuning when training data is sufficient, it tends to struggle in few-shot learning settings. In this paper, we present Multi-task Pre-trained Modular Prompt (MP2) to boost prompt tuning for few-shot learning. MP2 is a set of combinable prompts pre-trained on 38 Chinese tasks. On downstream tasks, the pre-trained prompts are selectively activated and combined, leading to strong compositional generalization to unseen tasks. To bridge the gap between pre-training and fine-tuning, we formulate upstream and downstream tasks into a unified machine reading comprehension task. Extensive experiments under two learning paradigms, i.e., gradient descent and black-box tuning, show that MP2 significantly outperforms prompt tuning, full model tuning, and prior prompt pre-training methods in few-shot settings. In addition, we demonstrate that MP2 can achieve surprisingly fast and strong adaptation to downstream tasks by merely learning 8 parameters to combine the pre-trained modular prompts.

COLO: A Contrastive Learning based Re-ranking Framework for One-Stage Summarization

Traditional training paradigms for extractive and abstractive summarization systems always only use token-level or sentence-level training objectives. However, the output summary is always evaluated from summary-level which leads to the inconsistency in training and evaluation. In this paper, we propose a Contrastive Learning based re-ranking framework for one-stage summarization called COLO. By modeling a contrastive objective, we show that the summarization model is able to directly generate summaries according to the summary-level score without additional modules and parameters. Extensive experiments demonstrate that COLO boosts the extractive and abstractive results of one-stage systems on CNN/DailyMail benchmark to 44.58 and 46.33 ROUGE-1 score while preserving the parameter efficiency and inference efficiency. Compared with state-of-the-art multi-stage systems, we save more than 100 GPU training hours and obtaining 3~8 speed-up ratio during inference while maintaining comparable results.