The value alignment of Large Language Models (LLMs) is critical because value is the foundation of LLM decision-making and behavior. Some recent work show that LLMs have similar value rankings. However, little is known about how susceptible LLM value rankings are to external influence and how different values are correlated with each other. In this work, we investigate the plasticity of LLM value systems by examining how their value rankings are influenced by different prompting strategies and exploring the intrinsic relationships between values. To this end, we design 6 different value transformation prompting methods including direct instruction, rubrics, in-context learning, scenario, persuasion, and persona, and benchmark the effectiveness of these methods on 3 different families and totally 8 LLMs. Our main findings include that the value rankings in large LLMs are much more susceptible to external influence than small LLMs, and there are intrinsic correlations between certain values (e.g., Privacy and Respect). Besides, through detailed correlation analysis, we find that the value correlations are more similar between large LLMs of different families than small LLMs of the same family. We also identify that scenario method is the strongest persuader and can help entrench the value rankings.

The rapid progress of reasoning and agentic large language models (LLMs) has increased the demand for long-context inference, but self-attention (SA) scales quadratically with context length. To address this, we propose SWARR (Sliding-Window Attention with Reinforced Adaptation for Math Reasoning), a practical recipe for adapting SWA models to mathematical reasoning. SWARR has two stages: (1) efficient conversion from a pretrained SA model to SWA with supervised fine-tuning (SFT), which avoids pretraining a new base model, and (2) architecture-aware policy adaptation with reinforcement learning (RL). We find that SWA still underperforms SA after SFT, and we hypothesize that this gap is caused in part by a data-architecture mismatch: most SFT data are prepared for SA models and may contain long-range dependencies that are difficult for SWA to model. Because RL is architecture-aware and optimizes self-generated trajectories under the SWA constraint, it can adapt training trajectories to better match the SWA architecture. Experiments on mathematical reasoning benchmarks show that this architecture-aware RL substantially narrows the gap between SWA and SA, recovering much of the accuracy lost during SWA conversion while preserving the efficiency benefits of linear-complexity attention.

In real-world multimodal applications, different modalities often exhibit weakly aligned adaptation requirements, making it challenging for a shared adapter to effectively capture heterogeneous signals. Existing approaches typically use a unified low-rank adapter for concatenated multimodal representations, leading to competition among modality-specific adaptations and diminishing returns as the adapter rank increases. To address this limitation, we propose msLoRA, a modality-specific adaptation framework for efficient multimodal fine-tuning. msLoRA assigns independent low-rank adapters to different modalities, allowing them to capture specialized adaptation directions while maintaining efficient parameterization. Our theoretical analysis introduces the concept of cumulative information span, demonstrating that modality-specific adapters can expand the effective adaptation space and avoid the cross-modal disagreement typical of shared adapters. We evaluate msLoRA using multiple prominent open-source LLMs, including the Llama, Qwen, and Mistral families, across four multimodal tasks, showing consistent improvements in scalability and effectiveness for low-rank adaptation in multimodal LLMs. The source code is available at https://github.com/XiuzeZhou/msLoRA/.

Individual investors are significantly outnumbered and disadvantaged in financial markets, overwhelmed by abundant information and lacking professional analysis. Equity research reports stand out as crucial resources, offering valuable insights. By leveraging these reports, large language models (LLMs) can enhance investors’ decision-making capabilities and strengthen financial analysis. However, two key challenges limit their effectiveness: (1) the rapid evolution of market events often outpaces the slow update cycles of existing knowledge bases, (2) the long-form and unstructured nature of financial reports further hinders timely and context-aware integration by LLMs. To address these challenges, we tackle both data and methodological aspects. First, we introduce the Event-Enhanced Automated Construction of Financial Knowledge Graph (FinKario), a dataset comprising over 305,360 entities, 210,328 relational triples, and 19 distinct relation types. FinKario automatically integrates real-time company fundamentals and market events through prompt-driven extraction guided by professional institutional templates, providing structured and accessible financial insights for LLMs. Additionally, we propose a Two-Stage, Graph-Based retrieval strategy (FinKario-RAG), optimizing the retrieval of evolving, large-scale financial knowledge to ensure efficient and precise data access. Extensive experiments show that FinKario with FinKario-RAG achieves superior stock trend prediction accuracy, outperforming financial LLMs by 18.81% and institutional strategies by 17.85% on average in backtesting. [Our code is available at https://github.com/Jackson906E/FinKario.]

Large language models (LLMs) face significant challenges in sustaining long-term memory for agentic applications due to limited context windows. To address this limitation, many work has proposed diverse memory mechanisms to support long-term, multi-turn interactions, leveraging different approaches tailored to distinct memory storage objects, such as KV caches. In this survey, we present a unified taxonomy that organizes memory systems for long-context scenarios by decoupling memory abstractions from model-specific inference and training methods. We categorize LLM memory into three primary paradigms: natural language tokens, intermediate representations and parameters. For each paradigm, we organize existing methods by three management stages, including memory construction, update, and query, so that long-context memory mechanisms can be described in a consistent way across system designs, with their implementation choices and constraints made explicit. Finally, we outline key research directions for long-context memory system design.

The rapid advancement of perovskite solar cells (PSCs) has led to an exponential growth in research publications, creating an urgent need for efficient knowledge management and reasoning systems in this domain. We present a comprehensive knowledge-enhanced system for PSCs that integrates three key components. First, we develop Perovskite-KG, a domain-specific knowledge graph constructed from 1,517 research papers, containing 23,789 entities and 22,272 relationships. Second, we create two complementary datasets: Perovskite-Chat, comprising 55,101 high-quality question-answer pairs generated through a novel multi-agent framework, and Perovskite-Reasoning, containing 2,217 carefully curated materials science problems. Third, we introduce two specialized large language models: Perovskite-Chat-LLM for domain-specific knowledge assistance and Perovskite-Reasoning-LLM for scientific reasoning tasks. Experimental results demonstrate that our system significantly outperforms existing models in both domain-specific knowledge retrieval and scientific reasoning tasks, providing researchers with effective tools for literature review, experimental design, and complex problem-solving in PSC research.

The creation of high-quality 3D scenes is essential for applications like video games and simulations, yet automating this process while retaining the benefits of Procedural Content Generation (PCG) remains challenging. In this paper, we introduce UnrealLLM, a novel multi-agent framework that connects natural language descriptions with the professional PCG system (Unreal Engine 5) to automate scene generation. UnrealLLM constructs a comprehensive knowledge base to translate text into executable PCG blueprints and a diverse asset library that guarantees high-quality scene generation. Additionally, it also introduces a text-based blueprint system with a spline-based control mechanism for geometric arrangement, enabling natural language interaction and enhancing interactivity in 3D environments using UE5’s advanced capabilities. Through extensive experiments, we show that UnrealLLM achieves competitive performance in technical metrics and aesthetic quality, offering unique advantages in generation scale and interactivity. This work makes a valuable contribution to automated 3D content creation, benefiting both novice users and professional designers.

Knowledge editing (KE) has emerged as a prominent alternative that enables efficient and precise information modification inside language models. However, a critical challenge arises in continuous language models editing — a significant performance decline both in knowledge update and retention when the number of edits increases. By dissecting the perturbation weight of language model in continuous KE, we uncover that disentangled and sparsified knowledge representation can significantly alleviate the performance decline. Building on these insights, we introduce AdaEdit, a novel knowledge editing method. Extensive empirical evaluations on multiple LLMs demonstrate that our proposed methods can enhance the performance of edited LLMs in large-size continuous editing regimes, outperforming existing ones without substantially compromising the general abilities of these models.

Despite the outstanding performance, Neural Architecture Search (NAS) is criticized for massive computation. Recently, Zero-shot NAS has emerged as a promising approach by exploiting Zero-cost (ZC) proxies, which markedly reduce computational demands. Despite this, existing ZC proxies heavily rely on expert knowledge and incur significant trial-and-error costs. Particularly in NLP tasks, most existing ZC proxies fail to surpass the performance of the naive baseline. To address these challenges, we introduce a novel framework, LPZero, which is the first to automatically design zero-cost (ZC) proxies for various tasks, achieving higher ranking consistency than human-designed proxies. Specifically, we model the ZC proxy as a symbolic equation and incorporate a unified proxy search space that encompasses existing ZC proxies, which are composed of a predefined set of mathematical symbols. To heuristically search for the best ZC proxy, LPZero incorporates genetic programming to find the optimal symbolic composition. We propose a Predictive-Pruning Strategy (PPS), which preemptively eliminates unpromising proxies, thereby mitigating the risk of proxy degradation. Extensive experiments on FlexiBERT, GPT-2, and LLaMA-7B demonstrate LPZero’s superior ranking ability and performance on downstream tasks compared to current approaches.

Current long-context benchmarks primarily focus on retrieval-based tests, requiring Large Language Models (LLMs) to locate specific information within extensive input contexts, such as the needle-in-a-haystack (NIAH) benchmark. Long-context generation refers to the ability of a language model to generate coherent and contextually accurate text that spans across lengthy passages or documents. While recent studies show strong performance on NIAH and other retrieval-based long-context benchmarks, there is a significant lack of benchmarks for evaluating long-context generation capabilities. To bridge this gap and offer a comprehensive assessment, we introduce a synthetic benchmark, LongGenBench, which allows for flexible configurations of customized generation context lengths. LongGenBench advances beyond traditional benchmarks by redesigning the format of questions and necessitating that LLMs respond with a single, cohesive long-context answer. Upon extensive evaluation using LongGenBench, we observe that: (1) both API accessed and open source models exhibit performance degradation in long-context generation scenarios, ranging from 1.2% to 47.1%; (2) different series of LLMs exhibit varying trends of performance degradation, with the Gemini-1.5-Flash model showing the least degradation among API accessed models, and the Qwen2 series exhibiting the least degradation in LongGenBench among open source models.

Model editing has become a promising method for precisely and effectively updating knowledge in language models. In this paper, we investigate knowledge attenuation, in which the retention of updated knowledge within the language model decreases as the number of edits increases after sequential editing. Through empirical study, we discovered that existing editing methods generally suffer from knowledge attenuation. We attribute this phenomenon to two aspects: (1) redundant parameters interference and (2) update weight disentanglement. To this end, we propose the AdaPLE method. It not only mitigates the knowledge attenuation issue but also improves the performance on existing benchmarks. To the best of our knowledge, we are the first to investigate the cause and mitigation of knowledge attenuation in sequential LLM editing.

The upscaling of Large Language Models (LLMs) has yielded impressive advances in natural language processing, yet it also poses significant deployment challenges. Weight quantization has emerged as a widely embraced solution to reduce memory and computational demands. This paper introduces BitDistiller, a framework that synergizes Quantization-Aware Training (QAT) with Knowledge Distillation (KD) to boost the performance of LLMs at ultra-low precisions (sub-4-bit). Specifically, BitDistiller first incorporates a tailored asymmetric quantization and clipping technique to maximally preserve the fidelity of quantized weights, and then proposes a novel Confidence-Aware Kullback-Leibler Divergence (CAKLD) objective, which is employed in a self-distillation manner to enable faster convergence and superior model performance. Empirical evaluations demonstrate that BitDistiller significantly surpasses existing methods in both 3-bit and 2-bit configurations on general language understanding and complex reasoning benchmarks. Notably, BitDistiller is shown to be more cost-effective, demanding fewer data and training resources. The code is available at https://github.com/DD-DuDa/BitDistiller.