How to Complete Domain Tuning while Keeping General Ability in LLM: Adaptive Layer-wise and Element-wise Regularization

Large Language Models (LLMs) exhibit strong general-purpose language capabilities. However, fine-tuning these models on domain-specific tasks often leads to catastrophic forgetting, where the model overwrites or loses essential knowledge acquired during pretraining. This phenomenon significantly limits the broader applicability of LLMs. To address this challenge, we propose a novel approach to compute the element-wise importance of model parameters crucial for preserving general knowledge during fine-tuning. Our method utilizes a dual-objective optimization strategy: (1) regularization loss to retain the parameter crucial for general knowledge; (2) cross-entropy loss to adapt to domain-specific tasks. Additionally, we introduce layer-wise coefficients to account for the varying contributions of different layers, dynamically balancing the dual-objective optimization. Extensive experiments on scientific, medical, and physical tasks using GPT-J and LLaMA-3 demonstrate that our approach mitigates catastrophic forgetting while enhancing model adaptability. Compared to previous methods, our solution is approximately 20 times faster and requires only 10%-15% of the storage, highlighting the practical efficiency. The code will be released.

SWSC: Shared Weight for Similar Channel in LLM

Large language models (LLMs) have spurred development in multiple industries. However, the growing number of their parameters brings substantial storage and computing burdens, making it essential to explore model compression techniques for parameter reduction and easier deployment. We propose SWSC, an LLM compression method based on the concept of Shared Weight for Similar Channel. It uses the K-Means clustering algorithm to cluster model weights channel-by-channel, generating clusters with highly similar vectors within each. A representative vector from each cluster is selected to approximately replace all vectors in the cluster, significantly reducing the number of model weight parameters. However, approximate restoration will inevitably cause damage to the performance of the model. To tackle this issue, we perform singular value decomposition on the weight error values before and after compression and retain the larger singular values and their corresponding singular vectors to compensate for the accuracy. The experimental results show that our method can effectively ensure the performance of the compressed LLM even under low-precision conditions.

DeepSeek-V3 Technical Report

We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) and DeepSeekMoE architectures, which were thoroughly validated in DeepSeek-V2. Furthermore, DeepSeek-V3 pioneers an auxiliary-loss-free strategy for load balancing and sets a multi-token prediction training objective for stronger performance. We pre-train DeepSeek-V3 on 14.8 trillion diverse and high-quality tokens, followed by Supervised Fine-Tuning and Reinforcement Learning stages to fully harness its capabilities. Comprehensive evaluations reveal that DeepSeek-V3 outperforms other open-source models and achieves performance comparable to leading closed-source models. Despite its excellent performance, DeepSeek-V3 requires only 2.788M H800 GPU hours for its full training. In addition, its training process is remarkably stable. Throughout the entire training process, we did not experience any irrecoverable loss spikes or perform any rollbacks. The model checkpoints are available at https://github.com/deepseek-ai/DeepSeek-V3.

LSAQ: Layer-Specific Adaptive Quantization for Large Language Model Deployment

As large language models (LLMs) demonstrate exceptional performance across various domains, the deployment of these models on edge devices has emerged as a new trend. Quantization techniques, which reduce the size and memory footprint of LLMs, are effective for enabling deployment on resource-constrained edge devices. However, existing one-size-fits-all quantization methods often fail to dynamically adjust the memory consumption of LLMs based on specific hardware characteristics and usage scenarios. To address this limitation, we propose LSAQ (Layer-Specific Adaptive Quantization), a system for adaptive quantization and dynamic deployment of LLMs based on layer importance. LSAQ evaluates layer importance by constructing top-k token sets from the inputs and outputs of each layer and calculating their Jaccard coefficient. Using this evaluation, the system adaptively adjusts quantization strategies in real time according to the resource availability of edge devices, assigning different precision levels to layers of varying importance. This approach significantly reduces the storage requirements of LLMs while maintaining model performance, enabling efficient deployment across diverse hardware platforms and usage scenarios.

BayLing 2: A Multilingual Large Language Model with Efficient Language Alignment

Large language models (LLMs), with their powerful generative capabilities and vast knowledge, empower various tasks in everyday life. However, these abilities are primarily concentrated in high-resource languages, leaving low-resource languages with weaker generative capabilities and relatively limited knowledge. Enhancing the multilingual capabilities of LLMs is therefore crucial for serving over 100 linguistic communities worldwide. An intuitive approach to enhance the multilingual capabilities would be to construct instruction data for various languages, but constructing instruction data for over 100 languages is prohibitively costly. In this paper, we introduce BayLing 2, which efficiently transfers generative capabilities and knowledge from high-resource languages to low-resource languages through language alignment. To achieve this, we constructed a dataset of 3.2 million instructions, comprising high-resource language instructions (Chinese and English) and cross-lingual instructions for 100+ languages and performed instruction tuning based on the dataset to facilitate the capability transfer between languages. Using Llama as the foundation model, we developed BayLing-2-7B, BayLing-2-13B, and BayLing-3-8B, and conducted a comprehensive evaluation of BayLing. For multilingual translation across 100+ languages, BayLing shows superior performance compared to open-source models of similar scale. For multilingual knowledge and understanding benchmarks, BayLing achieves significant improvements across over 20 low-resource languages, demonstrating its capability of effective knowledge transfer from high-resource to low-resource languages. Furthermore, results on English benchmarks indicate that BayLing maintains high performance in highresource languages while enhancing the performance in low-resource languages. Demo, homepage, code and models of BayLing are available.

Adaptively Controllable Diffusion Model for Efficient Conditional Image Generation

With the development of artificial intelligence, more and more attention has been put onto generative models, which represent the creativity, a very important aspect of intelligence. In recent years, diffusion models have been studied and proven to be more reasonable and effective than previous methods. However, common diffusion frameworks suffer from controllability problems. Although extra conditions have been considered by some work to guide the diffusion process for a specific target generation, it only controls the generation result but not its process. In this work, we propose a new adaptive framework, $\textit{Adaptively Controllable Diffusion (AC-Diff) Model}$, to automatically and fully control the generation process, including not only the type of generation result but also the length and parameters of the generation process. Both inputs and conditions will be first fed into a $\textit{Conditional Time-Step (CTS) Module}$ to determine the number of steps needed for a generation. Then according to the length of the process, the diffusion rate parameters will be estimated through our $\textit{Adaptive Hybrid Noise Schedule (AHNS) Module}$. We further train the network with the corresponding adaptive sampling mechanism to learn how to adjust itself according to the conditions for the overall performance improvement. To enable its practical applications, AC-Diff is expected to largely reduce the average number of generation steps and execution time while maintaining the same performance as done in the literature diffusion models.

Map-Free Trajectory Prediction with Map Distillation and Hierarchical Encoding

Reliable motion forecasting of surrounding agents is essential for ensuring the safe operation of autonomous vehicles. Many existing trajectory prediction methods rely heavily on high-definition (HD) maps as strong driving priors. However, the availability and accuracy of these priors are not guaranteed due to substantial costs to build, localization errors of vehicles, or ongoing road constructions. In this paper, we introduce MFTP, a Map-Free Trajectory Prediction method that offers several advantages. First, it eliminates the need for HD maps during inference while still benefiting from map priors during training via knowledge distillation. Second, we present a novel hierarchical encoder that effectively extracts spatial-temporal agent features and aggregates them into multiple trajectory queries. Additionally, we introduce an iterative decoder that sequentially decodes trajectory queries to generate the final predictions. Extensive experiments show that our approach achieves state-of-the-art performance on the Argoverse dataset under the map-free setting.

StreamAdapter: Efficient Test Time Adaptation from Contextual Streams

In-context learning (ICL) allows large language models (LLMs) to adapt to new tasks directly from the given demonstrations without requiring gradient updates. While recent advances have expanded context windows to accommodate more demonstrations, this approach increases inference costs without necessarily improving performance. To mitigate these issues, We propose StreamAdapter, a novel approach that directly updates model parameters from context at test time, eliminating the need for explicit in-context demonstrations. StreamAdapter employs context mapping and weight absorption mechanisms to dynamically transform ICL demonstrations into parameter updates with minimal additional parameters. By reducing reliance on numerous in-context examples, StreamAdapter significantly reduce inference costs and allows for efficient inference with constant time complexity, regardless of demonstration count. Extensive experiments across diverse tasks and model architectures demonstrate that StreamAdapter achieves comparable or superior adaptation capability to ICL while requiring significantly fewer demonstrations. The superior task adaptation and context encoding capabilities of StreamAdapter on both language understanding and generation tasks provides a new perspective for adapting LLMs at test time using context, allowing for more efficient adaptation across scenarios and more cost-effective inference

Model Editing for LLMs4Code: How Far are We?

Large Language Models for Code (LLMs4Code) have been found to exhibit outstanding performance in the software engineering domain, especially the remarkable performance in coding tasks. However, even the most advanced LLMs4Code can inevitably contain incorrect or outdated code knowledge. Due to the high cost of training LLMs4Code, it is impractical to re-train the models for fixing these problematic code knowledge. Model editing is a new technical field for effectively and efficiently correcting erroneous knowledge in LLMs, where various model editing techniques and benchmarks have been proposed recently. Despite that, a comprehensive study that thoroughly compares and analyzes the performance of the state-of-the-art model editing techniques for adapting the knowledge within LLMs4Code across various code-related tasks is notably absent. To bridge this gap, we perform the first systematic study on applying state-of-the-art model editing approaches to repair the inaccuracy of LLMs4Code. To that end, we introduce a benchmark named CLMEEval, which consists of two datasets, i.e., CoNaLa-Edit (CNLE) with 21K+ code generation samples and CodeSearchNet-Edit (CSNE) with 16K+ code summarization samples. With the help of CLMEEval, we evaluate six advanced model editing techniques on three LLMs4Code: CodeLlama (7B), CodeQwen1.5 (7B), and Stable-Code (3B). Our findings include that the external memorization-based GRACE approach achieves the best knowledge editing effectiveness and specificity (the editing does not influence untargeted knowledge), while generalization (whether the editing can generalize to other semantically-identical inputs) is a universal challenge for existing techniques. Furthermore, building on in-depth case analysis, we introduce an enhanced version of GRACE called A-GRACE, which incorporates contrastive learning to better capture the semantics of the inputs.

Generative Adapter: Contextualizing Language Models in Parameters with A Single Forward Pass

Large language models (LMs) are typically adapted to improve performance on new contexts (\eg text prompts that define new tasks or domains) through fine-tuning or prompting. However, there is an accuracy compute tradeoff -- fine-tuning incurs significant training cost and prompting increases inference overhead. We introduce $GenerativeAdapter$, an effective and efficient adaptation method that directly maps new contexts to low-rank LM adapters, thereby significantly reducing inference overhead with no need for finetuning. The adapter generator is trained via self-supervised learning, and can be used to adapt a single frozen LM for any new task simply by mapping the associated task or domain context to a new adapter. We apply $GenerativeAdapter$ to two pretrained LMs (Mistral-7B-Instruct and Llama2-7B-Chat) and evaluate the adapted models in three adaption scenarios: knowledge acquisition from documents, learning from demonstrations, and personalization for users. In StreamingQA, our approach is effective in injecting knowledge into the LM's parameters, achieving a 63.5% improvement in F1 score over the model with supervised fine-tuning (from $19.5$ to $31.5$) for contexts as long as 32K tokens. In the MetaICL in-context learning evaluation, our method achieves an average accuracy of $44.9$ across 26 tasks, outperforming the base model. On MSC, our method proves to be highly competitive in memorizing user information from conversations with a 4x reduction in computation and memory costs compared to prompting with full conversation history. Together, these results suggest that $GenerativeAdapter$ should allow for general adaption to a wide range of different contexts.