Synthetic Data Generation & Multi-Step RL for Reasoning & Tool Use

Reinforcement learning has been shown to improve the performance of large language models. However, traditional approaches like RLHF or RLAIF treat the problem as single-step. As focus shifts toward more complex reasoning and agentic tasks, language models must take multiple steps of text generation, reasoning and environment interaction before generating a solution. We propose a synthetic data generation and RL methodology targeting multi-step optimization scenarios. This approach, called Step-Wise Reinforcement Learning (SWiRL), iteratively generates multi-step reasoning and tool use data, and then learns from that data. It employs a simple step-wise decomposition that breaks each multi-step trajectory into multiple sub-trajectories corresponding to each action by the original model. It then applies synthetic data filtering and RL optimization on these sub-trajectories. We evaluated SWiRL on a number of multi-step tool use, question answering, and mathematical reasoning tasks. Our experiments show that SWiRL outperforms baseline approaches by 21.5%, 12.3%, 14.8%, 11.1%, and 15.3% in relative accuracy on GSM8K, HotPotQA, CofCA, MuSiQue, and BeerQA, respectively. Excitingly, the approach exhibits generalization across tasks: for example, training only on HotPotQA (text question-answering) improves zero-shot performance on GSM8K (a math dataset) by a relative 16.9%.

Investigating and Scaling up Code-Switching for Multilingual Language Model Pre-Training

Large language models (LLMs) exhibit remarkable multilingual capabilities despite the extreme language imbalance in the pre-training data. In this paper, we closely examine the reasons behind this phenomenon, focusing on the pre-training corpus. We find that the existence of code-switching, alternating between different languages within a context, is key to multilingual capabilities. We conduct an analysis to investigate code-switching in the pre-training corpus, examining its presence and categorizing it into four types within two quadrants. We then assess its impact on multilingual performance. These types of code-switching data are unbalanced in proportions and demonstrate different effects on facilitating language transfer. To better explore the power of code-switching for language alignment during pre-training, we investigate the strategy of synthetic code-switching. We continuously scale up the synthetic code-switching data and observe remarkable improvements in both benchmarks and representation space. Extensive experiments indicate that incorporating synthetic code-switching data enables better language alignment and generalizes well to high, medium, and low-resource languages with pre-training corpora of varying qualities.

Enhancing Large Language Models (LLMs) for Telecommunications using Knowledge Graphs and Retrieval-Augmented Generation

Large language models (LLMs) have made significant progress in general-purpose natural language processing tasks. However, LLMs are still facing challenges when applied to domain-specific areas like telecommunications, which demands specialized expertise and adaptability to evolving standards. This paper presents a novel framework that combines knowledge graph (KG) and retrieval-augmented generation (RAG) techniques to enhance LLM performance in the telecom domain. The framework leverages a KG to capture structured, domain-specific information about network protocols, standards, and other telecom-related entities, comprehensively representing their relationships. By integrating KG with RAG, LLMs can dynamically access and utilize the most relevant and up-to-date knowledge during response generation. This hybrid approach bridges the gap between structured knowledge representation and the generative capabilities of LLMs, significantly enhancing accuracy, adaptability, and domain-specific comprehension. Our results demonstrate the effectiveness of the KG-RAG framework in addressing complex technical queries with precision. The proposed KG-RAG model attained an accuracy of 88% for question answering tasks on a frequently used telecom-specific dataset, compared to 82% for the RAG-only and 48% for the LLM-only approaches.

Change3D: Revisiting Change Detection and Captioning from A Video Modeling Perspective

In this paper, we present Change3D, a framework that reconceptualizes the change detection and captioning tasks through video modeling. Recent methods have achieved remarkable success by regarding each pair of bi-temporal images as separate frames. They employ a shared-weight image encoder to extract spatial features and then use a change extractor to capture differences between the two images. However, image feature encoding, being a task-agnostic process, cannot attend to changed regions effectively. Furthermore, different change extractors designed for various change detection and captioning tasks make it difficult to have a unified framework. To tackle these challenges, Change3D regards the bi-temporal images as comprising two frames akin to a tiny video. By integrating learnable perception frames between the bi-temporal images, a video encoder enables the perception frames to interact with the images directly and perceive their differences. Therefore, we can get rid of the intricate change extractors, providing a unified framework for different change detection and captioning tasks. We verify Change3D on multiple tasks, encompassing change detection (including binary change detection, semantic change detection, and building damage assessment) and change captioning, across eight standard benchmarks. Without bells and whistles, this simple yet effective framework can achieve superior performance with an ultra-light video model comprising only ~6%-13% of the parameters and ~8%-34% of the FLOPs compared to state-of-the-art methods. We hope that Change3D could be an alternative to 2D-based models and facilitate future research.

DAPO: An Open-Source LLM Reinforcement Learning System at Scale

Inference scaling empowers LLMs with unprecedented reasoning ability, with reinforcement learning as the core technique to elicit complex reasoning. However, key technical details of state-of-the-art reasoning LLMs are concealed (such as in OpenAI o1 blog and DeepSeek R1 technical report), thus the community still struggles to reproduce their RL training results. We propose the $\textbf{D}$ecoupled Clip and $\textbf{D}$ynamic s$\textbf{A}$mpling $\textbf{P}$olicy $\textbf{O}$ptimization ($\textbf{DAPO}$) algorithm, and fully open-source a state-of-the-art large-scale RL system that achieves 50 points on AIME 2024 using Qwen2.5-32B base model. Unlike previous works that withhold training details, we introduce four key techniques of our algorithm that make large-scale LLM RL a success. In addition, we open-source our training code, which is built on the verl framework, along with a carefully curated and processed dataset. These components of our open-source system enhance reproducibility and support future research in large-scale LLM RL.

MACS: Multi-source Audio-to-image Generation with Contextual Significance and Semantic Alignment

Propelled by the breakthrough in deep generative models, audio-to-image generation has emerged as a pivotal cross-model task that converts complex auditory signals into rich visual representations. However, previous works only focus on single-source audio inputs for image generation, ignoring the multi-source characteristic in natural auditory scenes, thus limiting the performance in generating comprehensive visual content. To bridge this gap, a method called MACS is proposed to conduct multi-source audio-to-image generation. This is the first work that explicitly separates multi-source audio to capture the rich audio components before image generation. MACS is a two-stage method. In the first stage, multi-source audio inputs are separated by a weakly supervised method, where the audio and text labels are semantically aligned by casting into a common space using the large pre-trained CLAP model. We introduce a ranking loss to consider the contextual significance of the separated audio signals. In the second stage, efficient image generation is achieved by mapping the separated audio signals to the generation condition using only a trainable adapter and a MLP layer. We preprocess the LLP dataset as the first full multi-source audio-to-image generation benchmark. The experiments are conducted on multi-source, mixed-source, and single-source audio-to-image generation tasks. The proposed MACS outperforms the current state-of-the-art methods in 17 of the 21 evaluation indexes on all tasks and delivers superior visual quality. The code will be publicly available.

Assessing the Macro and Micro Effects of Random Seeds on Fine-Tuning Large Language Models

The impact of random seeds in fine-tuning large language models (LLMs) has been largely overlooked despite its potential influence on model performance.In this study, we systematically evaluate the effects of random seeds on LLMs using the GLUE and SuperGLUE benchmarks. We analyze the macro-level impact through traditional metrics like accuracy and F1, calculating their mean and variance to quantify performance fluctuations. To capture the micro-level effects, we introduce a novel metric, consistency, measuring the stability of individual predictions across runs. Our experiments reveal significant variance at both macro and micro levels, underscoring the need for careful consideration of random seeds in fine-tuning and evaluation.

Erase Diffusion: Empowering Object Removal Through Calibrating Diffusion Pathways

Erase inpainting, or object removal, aims to precisely remove target objects within masked regions while preserving the overall consistency of the surrounding content. Despite diffusion-based methods have made significant strides in the field of image inpainting, challenges remain regarding the emergence of unexpected objects or artifacts. We assert that the inexact diffusion pathways established by existing standard optimization paradigms constrain the efficacy of object removal. To tackle these challenges, we propose a novel Erase Diffusion, termed EraDiff, aimed at unleashing the potential power of standard diffusion in the context of object removal. In contrast to standard diffusion, the EraDiff adapts both the optimization paradigm and the network to improve the coherence and elimination of the erasure results. We first introduce a Chain-Rectifying Optimization (CRO) paradigm, a sophisticated diffusion process specifically designed to align with the objectives of erasure. This paradigm establishes innovative diffusion transition pathways that simulate the gradual elimination of objects during optimization, allowing the model to accurately capture the intent of object removal. Furthermore, to mitigate deviations caused by artifacts during the sampling pathways, we develop a simple yet effective Self-Rectifying Attention (SRA) mechanism. The SRA calibrates the sampling pathways by altering self-attention activation, allowing the model to effectively bypass artifacts while further enhancing the coherence of the generated content. With this design, our proposed EraDiff achieves state-of-the-art performance on the OpenImages V5 dataset and demonstrates significant superiority in real-world scenarios.

MoMa: A Modular Deep Learning Framework for Material Property Prediction

Deep learning methods for material property prediction have been widely explored to advance materials discovery. However, the prevailing pre-train then fine-tune paradigm often fails to address the inherent diversity and disparity of material tasks. To overcome these challenges, we introduce MoMa, a Modular framework for Materials that first trains specialized modules across a wide range of tasks and then adaptively composes synergistic modules tailored to each downstream scenario. Evaluation across 17 datasets demonstrates the superiority of MoMa, with a substantial 14% average improvement over the strongest baseline. Few-shot and continual learning experiments further highlight MoMa's potential for real-world applications. Pioneering a new paradigm of modular material learning, MoMa will be open-sourced to foster broader community collaboration.

FitLight: Federated Imitation Learning for Plug-and-Play Autonomous Traffic Signal Control

Although Reinforcement Learning (RL)-based Traffic Signal Control (TSC) methods have been extensively studied, their practical applications still raise some serious issues such as high learning cost and poor generalizability. This is because the ``trial-and-error'' training style makes RL agents extremely dependent on the specific traffic environment, which also requires a long convergence time. To address these issues, we propose a novel Federated Imitation Learning (FIL)-based framework for multi-intersection TSC, named FitLight, which allows RL agents to plug-and-play for any traffic environment without additional pre-training cost. Unlike existing imitation learning approaches that rely on pre-training RL agents with demonstrations, FitLight allows real-time imitation learning and seamless transition to reinforcement learning. Due to our proposed knowledge-sharing mechanism and novel hybrid pressure-based agent design, RL agents can quickly find a best control policy with only a few episodes. Moreover, for resource-constrained TSC scenarios, FitLight supports model pruning and heterogeneous model aggregation, such that RL agents can work on a micro-controller with merely 16{\it KB} RAM and 32{\it KB} ROM. Extensive experiments demonstrate that, compared to state-of-the-art methods, FitLight not only provides a superior starting point but also converges to a better final solution on both real-world and synthetic datasets, even under extreme resource limitations.