Knowledge distillation (KD) is a promising solution to compress large language models (LLMs) by transferring their knowledge to smaller models. During this process, white-box KD methods usually minimize the distance between the output distributions of the teacher model and the student model to transfer more information. However, we reveal that the current white-box KD framework exhibits two limitations: a) bridging probability distributions from different output spaces will limit the similarity between the teacher model and the student model; b) this framework cannot be applied to LLMs with different vocabularies. One of the root causes for these limitations is that the distributions from the teacher and the student for KD are output by different prediction heads, which yield distributions in different output spaces and dimensions. Therefore, in this paper, we propose a dual-space knowledge distillation (DSKD) framework that unifies the prediction heads of the teacher and the student models for KD. Specifically, we first introduce two projectors with ideal initialization to project the teacher/student hidden states into the student/teacher representation spaces. After this, the hidden states from different models can share the same head and unify the output spaces of the distributions. Furthermore, we develop an exact token alignment (ETA) algorithm to align the same tokens in two differently-tokenized sequences. Based on the above, our DSKD framework is a general KD framework that supports both off-policy and on-policy KD, and KD between any two LLMs regardless of their vocabularies. Extensive experiments on instruction-following, mathematical reasoning, and code generation benchmarks show that DSKD significantly outperforms existing methods based on the current white-box KD framework and surpasses other cross-tokenizer KD methods for LLMs with different vocabularies.
Generating high-quality code that solves complex programming tasks is challenging, especially with current decoder-based models that produce highly stochastic outputs. In code generation, even minor errors can easily break the entire solution. Leveraging multiple sampled solutions can significantly improve the overall output quality. One effective way to enhance code generation is by pairing a code generation model with a reranker model, which selects the best solution from the generated samples. We propose a novel iterative self-training approach for self-training reranker models using Proximal Policy Optimization (PPO), aimed at improving both reranking accuracy and the overall code generation process. Unlike traditional PPO approaches, where the focus is on optimizing a generative model with a reward model, our approach emphasizes the development of a robust reward/reranking model. This model improves the quality of generated code through reranking and addresses problems and errors that the reward model might overlook during PPO alignment with the reranker. Our method iteratively refines the training dataset by re-evaluating outputs, identifying high-scoring negative examples, and incorporating them into the training loop, that boosting model performance. Our evaluation on the MultiPL-E dataset demonstrates that our 13.4B parameter model outperforms a 33B model in code generation quality while being three times faster. Moreover, it achieves performance comparable to GPT-4 and surpasses it in one programming language.
Large language models (LLMs) have achieved notable success in code generation. However, they still frequently produce uncompilable output because their next-token inference procedure does not model formal aspects of code. Although constrained decoding is a promising approach to alleviate this issue, it has only been applied to handle either domain-specific languages or syntactic features of general-purpose programming languages. However, LLMs frequently generate code with typing errors, which are beyond the domain of syntax and generally hard to adequately constrain. To address this challenge, we introduce a type-constrained decoding approach that leverages type systems to guide code generation. For this purpose, we develop novel prefix automata and a search over inhabitable types, forming a sound approach to enforce well-typedness on LLM-generated code. We formalize our approach on a foundational simply-typed language and extend it to TypeScript to demonstrate practicality. Our evaluation on the HumanEval and MBPP datasets shows that our approach reduces compilation errors by more than half and significantly increases functional correctness in code synthesis, translation, and repair tasks across LLMs of various sizes and model families, including state-of-the-art open-weight models with more than 30B parameters. The results demonstrate the generality and effectiveness of our approach in constraining LLM code generation with formal rules of type systems.
Multimodal Large Language Models (MLLMs) are set to transform how machines process and generate human-like responses by integrating diverse modalities such as text, images, and code. Yet, effectively harnessing their capabilities hinges on optimal prompt engineering. We present a comprehensive experimental evaluation of seven prompt engineering methods applied to 13 open-source MLLMs over 24 tasks spanning Reasoning and Compositionality, Multimodal Understanding and Alignment, Complex Code Generation and Execution, and Knowledge Retrieval and Integration. Our approach stratifies models by parameter count into Small (<4B), Medium (4B-10B), and Large (>10B) categories and compares prompting techniques including Zero-Shot, One-Shot, Few-Shot, Chain-of-Thought, Analogical, Generated Knowledge, and Tree-of-Thought. While Large MLLMs excel in structured tasks such as code generation, achieving accuracies up to 96.88% under Few-Shot prompting, all models struggle with complex reasoning and abstract understanding, often yielding accuracies below 60% and high hallucination rates. Structured reasoning prompts frequently increased hallucination up to 75% in small models and led to longer response times (over 20 seconds in Large MLLMs), while simpler prompting methods provided more concise and efficient outputs. No single prompting method uniformly optimises all task types. Instead, adaptive strategies combining example-based guidance with selective structured reasoning are essential to enhance robustness, efficiency, and factual accuracy. Our findings offer practical recommendations for prompt engineering and support more reliable deployment of MLLMs across applications including AI-assisted coding, knowledge retrieval, and multimodal content understanding.
The emergence of large language models (LLMs) has significantly promoted the development of code generation task, sparking a surge in pertinent literature. Current research is hindered by redundant generation results and a tendency to overfit local patterns in the short term. Although existing studies attempt to alleviate the issue by adopting a multi-token prediction strategy, there remains limited focus on choosing the appropriate processing length for generations. By analyzing the attention between tokens during the generation process of LLMs, it can be observed that the high spikes of the attention scores typically appear at the end of lines. This insight suggests that it is reasonable to treat each line of code as a fundamental processing unit and generate them sequentially. Inspired by this, we propose the \textbf{LSR-MCTS} algorithm, which leverages MCTS to determine the code line-by-line and select the optimal path. Further, we integrate a self-refine mechanism at each node to enhance diversity and generate higher-quality programs through error correction. Extensive experiments and comprehensive analyses on three public coding benchmarks demonstrate that our method outperforms the state-of-the-art performance approaches.
Large Language Models (LLMs) are emerging as promising tools for automated reinforcement learning (RL) reward design, owing to their robust capabilities in commonsense reasoning and code generation. By engaging in dialogues with RL agents, LLMs construct a Reward Observation Space (ROS) by selecting relevant environment states and defining their internal operations. However, existing frameworks have not effectively leveraged historical exploration data or manual task descriptions to iteratively evolve this space. In this paper, we propose a novel heuristic framework that enhances LLM-driven reward design by evolving the ROS through a table-based exploration caching mechanism and a text-code reconciliation strategy. Our framework introduces a state execution table, which tracks the historical usage and success rates of environment states, overcoming the Markovian constraint typically found in LLM dialogues and facilitating more effective exploration. Furthermore, we reconcile user-provided task descriptions with expert-defined success criteria using structured prompts, ensuring alignment in reward design objectives. Comprehensive evaluations on benchmark RL tasks demonstrate the effectiveness and stability of the proposed framework. Code and video demos are available at jingjjjjjie.github.io/LLM2Reward.
As Large Language Models (LLMs) are increasingly applied across various tasks, instruction tuning has emerged as a critical method for enhancing model performance. However, current data management strategies face substantial challenges in generating diverse and comprehensive data, restricting further improvements in model performance. To address this gap, we propose MDIT, a novel model-free data interpolation method for diverse instruction tuning, which generates varied and high-quality instruction data by performing task interpolation. Moreover, it contains diversity-based clustering strategies to ensure the diversity of the training data. Extensive experiments show that our method achieves superior performance in multiple benchmark tasks. The LLMs finetuned with MDIT show significant improvements in numerous tasks such as general question answering, math reasoning, and code generation. MDIT offers an efficient and automatic data synthetic method, generating diverse instruction data without depending on external resources while expanding the application potential of LLMs in complex environments.
Large Language Models (LLMs) have demonstrated remarkable capabilities in code generation. However, the quality of the generated code is heavily dependent on the structure and composition of the prompts used. Crafting high-quality prompts is a challenging task that requires significant knowledge and skills of prompt engineering. To advance the automation support for the prompt engineering for LLM-based code generation, we propose a novel solution Diffusion-Driven Prompt Tuning (DDPT) that learns how to generate optimal prompt embedding from Gaussian Noise to automate the prompt engineering for code generation. We evaluate the feasibility of diffusion-based optimization and abstract the optimal prompt embedding as a directional vector toward the optimal embedding. We use the code generation loss given by the LLMs to help the diffusion model capture the distribution of optimal prompt embedding during training. The trained diffusion model can build a path from the noise distribution to the optimal distribution at the sampling phrase, the evaluation result demonstrates that DDPT helps improve the prompt optimization for code generation.
Large Language Models (LLMs) have transformed software development by enabling code generation, automated debugging, and complex reasoning. However, their continued advancement is constrained by the scarcity of high-quality, publicly available supervised fine-tuning (SFT) datasets tailored for coding tasks. To bridge this gap, we introduce OpenCodeInstruct, the largest open-access instruction tuning dataset, comprising 5 million diverse samples. Each sample includes a programming question, solution, test cases, execution feedback, and LLM-generated quality assessments. We fine-tune various base models, including LLaMA and Qwen, across multiple scales (1B+, 3B+, and 7B+) using our dataset. Comprehensive evaluations on popular benchmarks (HumanEval, MBPP, LiveCodeBench, and BigCodeBench) demonstrate substantial performance improvements achieved by SFT with OpenCodeInstruct. We also present a detailed methodology encompassing seed data curation, synthetic instruction and solution generation, and filtering.
The rapid advancement of Large Language Models (LLMs) has outpaced traditional evaluation methods. Static benchmarks fail to capture the depth and breadth of LLM capabilities and eventually become obsolete, while most dynamic approaches either rely too heavily on LLM-based evaluation or remain constrained by predefined test sets. We introduce Prism, a flexible, dynamic benchmarking framework designed for comprehensive LLM assessment. Prism builds on three key components: (1) a tree-based state representation that models evaluation as a Markov Decision Process, (2) a Monte Carlo Tree Search algorithm adapted to uncover challenging evaluation scenarios, and (3) a multi-agent evaluation pipeline that enables simultaneous assessment of diverse capabilities. To ensure robust evaluation, Prism integrates structural measurements of tree exploration patterns with performance metrics across difficulty levels, providing detailed diagnostics of error patterns, test coverage, and solution approaches. Through extensive experiments on five state-of-the-art LLMs, we analyze how model architecture and scale influence code generation performance across varying task difficulties. Our results demonstrate Prism's effectiveness as a dynamic benchmark that evolves with model advancements while offering deeper insights into their limitations.
Large Language Models (LLMs) are increasingly used in post-development tasks such as code repair and testing. A key factor in these tasks' success is the model's deep understanding of code. However, the extent to which LLMs truly understand code remains largely unevaluated. Quantifying code comprehension is challenging due to its abstract nature and the lack of a standardized metric. Previously, this was assessed through developer surveys, which are not feasible for evaluating LLMs. Existing LLM benchmarks focus primarily on code generation, fundamentally different from code comprehension. Additionally, fixed benchmarks quickly become obsolete as they become part of the training data. This paper presents the first large-scale empirical investigation into LLMs' ability to understand code. Inspired by mutation testing, we use an LLM's fault-finding ability as a proxy for its deep code understanding. This approach is based on the insight that a model capable of identifying subtle functional discrepancies must understand the code well. We inject faults in real-world programs and ask the LLM to localize them, ensuring the specifications suffice for fault localization. Next, we apply semantic-preserving code mutations (SPMs) to the faulty programs and test whether the LLMs still locate the faults, verifying their confidence in code understanding. We evaluate nine popular LLMs on 600,010 debugging tasks from 670 Java and 637 Python programs. We find that LLMs lose the ability to debug the same bug in 78% of faulty programs when SPMs are applied, indicating a shallow understanding of code and reliance on features irrelevant to semantics. We also find that LLMs understand code earlier in the program better than later. This suggests that LLMs' code comprehension remains tied to lexical and syntactic features due to tokenization designed for natural languages, which overlooks code semantics.
Large Language Models (LLMs), such as GPT-4 and DeepSeek, have been applied to a wide range of domains in software engineering. However, their potential in the context of High-Performance Computing (HPC) much remains to be explored. This paper evaluates how well DeepSeek, a recent LLM, performs in generating a set of HPC benchmark codes: a conjugate gradient solver, the parallel heat equation, parallel matrix multiplication, DGEMM, and the STREAM triad operation. We analyze DeepSeek's code generation capabilities for traditional HPC languages like Cpp, Fortran, Julia and Python. The evaluation includes testing for code correctness, performance, and scaling across different configurations and matrix sizes. We also provide a detailed comparison between DeepSeek and another widely used tool: GPT-4. Our results demonstrate that while DeepSeek generates functional code for HPC tasks, it lags behind GPT-4, in terms of scalability and execution efficiency of the generated code.
Large Language Models (LLMs) have demonstrated remarkable performance across a wide range of applications, e.g., medical question-answering, mathematical sciences, and code generation. However, they also exhibit inherent limitations, such as outdated knowledge and susceptibility to hallucinations. Retrieval-Augmented Generation (RAG) has emerged as a promising paradigm to address these issues, but it also introduces new vulnerabilities. Recent efforts have focused on the security of RAG-based LLMs, yet existing attack methods face three critical challenges: (1) their effectiveness declines sharply when only a limited number of poisoned texts can be injected into the knowledge database, (2) they lack sufficient stealth, as the attacks are often detectable by anomaly detection systems, which compromises their effectiveness, and (3) they rely on heuristic approaches to generate poisoned texts, lacking formal optimization frameworks and theoretic guarantees, which limits their effectiveness and applicability. To address these issues, we propose coordinated Prompt-RAG attack (PR-attack), a novel optimization-driven attack that introduces a small number of poisoned texts into the knowledge database while embedding a backdoor trigger within the prompt. When activated, the trigger causes the LLM to generate pre-designed responses to targeted queries, while maintaining normal behavior in other contexts. This ensures both high effectiveness and stealth. We formulate the attack generation process as a bilevel optimization problem leveraging a principled optimization framework to develop optimal poisoned texts and triggers. Extensive experiments across diverse LLMs and datasets demonstrate the effectiveness of PR-Attack, achieving a high attack success rate even with a limited number of poisoned texts and significantly improved stealth compared to existing methods.
Low-Rank Adaptation (LoRA) has emerged as a popular parameter-efficient fine-tuning (PEFT) method for Large Language Models (LLMs), yet it still incurs notable overhead and suffers from parameter interference in multi-task scenarios. We propose LoRA with Reduced Interference (LoRI), a simple yet effective approach that freezes the projection matrices $A$ as random projections and sparsifies the matrices $B$ using task-specific masks. This design substantially reduces the number of trainable parameters while maintaining strong task performance. Moreover, LoRI minimizes cross-task interference in adapter merging by leveraging the orthogonality between adapter subspaces, and supports continual learning by using sparsity to mitigate catastrophic forgetting. Extensive experiments across natural language understanding, mathematical reasoning, code generation, and safety alignment tasks demonstrate that LoRI outperforms full fine-tuning and existing PEFT methods, while using up to 95% fewer trainable parameters than LoRA. In multi-task experiments, LoRI enables effective adapter merging and continual learning with reduced cross-task interference. Code is available at: https://github.com/juzhengz/LoRI
Chart-to-code generation, the process of converting chart images into executable plotting scripts, provides a lossless representation of chart information, requiring models to accurately capture and summarize all visual and structural elements. However, this remains a significant challenge for multimodal large language models (MLLMs), which are not inherently well-aligned with code generation tasks. To bridge this gap, we introduce Chart2Code, a novel iterative dual preference learning framework designed to enhance MLLMs' chart-to-code generation capabilities through structured code variant generation and fine-grained dual reward signals. We validate Chart2Code across three MLLMs and find that iterative preference learning consistently improves out-of-distribution chart-to-code generation quality. Throughout this process, our dual scoring method, which evaluates both the textual code structure and its visual representation, leads to greater performance improvements, even with a reduced preference dataset size. Further analysis explores the key components of our framework and highlights the interplay between chart-to-code generation and broader chart reasoning, paving the way for future advancements in chart comprehension.
This paper introduces a self-learning agent that integrates LLaMA 3.2 with a Progressive Neural Network (PNN) for continual learning in conversational AI and code generation. The framework dynamically collects data, fine-tunes tasks with minimal samples, and leverages Meta-Learning for rapid adaptation. LoRA optimizes fine-tuning, while Elastic Weight Consolidation (EWC) enhances knowledge retention. Experimental results demonstrate improved adaptability and memory stability, positioning this approach as a scalable step toward Artificial General Intelligence (AGI).
Automated Driving System (ADS) is a safety-critical software system responsible for the interpretation of the vehicle's environment and making decisions accordingly. The unbounded complexity of the driving context, including unforeseeable events, necessitate continuous improvement, often achieved through iterative DevOps processes. However, DevOps processes are themselves complex, making these improvements both time- and resource-intensive. Automation in code generation for ADS using Large Language Models (LLM) is one potential approach to address this challenge. Nevertheless, the development of ADS requires rigorous processes to verify, validate, assess, and qualify the code before it can be deployed in the vehicle and used. In this study, we developed and evaluated a prototype for automatic code generation and assessment using a designed pipeline of a LLM-based agent, simulation model, and rule-based feedback generator in an industrial setup. The LLM-generated code is evaluated automatically in a simulation model against multiple critical traffic scenarios, and an assessment report is provided as feedback to the LLM for modification or bug fixing. We report about the experimental results of the prototype employing Codellama:34b, DeepSeek (r1:32b and Coder:33b), CodeGemma:7b, Mistral:7b, and GPT4 for Adaptive Cruise Control (ACC) and Unsupervised Collision Avoidance by Evasive Manoeuvre (CAEM). We finally assessed the tool with 11 experts at two Original Equipment Manufacturers (OEMs) by conducting an interview study.
General image-to-video generation methods often produce suboptimal animations that do not meet the requirements of animated graphics, as they lack active text motion and exhibit object distortion. Also, code-based animation generation methods typically require layer-structured vector data which are often not readily available for motion graphic generation. To address these challenges, we propose a novel framework named MG-Gen that reconstructs data in vector format from a single raster image to extend the capabilities of code-based methods to enable motion graphics generation from a raster image in the framework of general image-to-video generation. MG-Gen first decomposes the input image into layer-wise elements, reconstructs them as HTML format data and then generates executable JavaScript code for the reconstructed HTML data. We experimentally confirm that MG-Gen generates motion graphics while preserving text readability and input consistency. These successful results indicate that combining layer decomposition and animation code generation is an effective strategy for motion graphics generation.
The rapid advancements in LLMs have driven the adoption of generative AI in various domains, including Electronic Design Automation (EDA). Unlike traditional software development, EDA presents unique challenges, as generated RTL code must not only be syntactically correct and functionally accurate but also synthesizable by hardware generators while meeting performance, power, and area constraints. These additional requirements introduce complexities that existing code-generation benchmarks often fail to capture, limiting their effectiveness in evaluating LLMs for RTL generation. To address this gap, we propose TuRTLe, a unified evaluation framework designed to systematically assess LLMs across key RTL generation tasks. TuRTLe integrates multiple existing benchmarks and automates the evaluation process, enabling a comprehensive assessment of LLM performance in syntax correctness, functional correctness, synthesis, PPA optimization, and exact line completion. Using this framework, we benchmark a diverse set of open LLMs and analyze their strengths and weaknesses in EDA-specific tasks. Our results show that reasoning-based models, such as DeepSeek R1, consistently outperform others across multiple evaluation criteria, but at the cost of increased computational overhead and inference latency. Additionally, base models are better suited in module completion tasks, while instruct-tuned models perform better in specification-to-RTL tasks.
Large language models (LLMs) have revolutionized code generation, significantly enhancing developer productivity. However, for a vast number of users with minimal coding knowledge, LLMs provide little support, as they primarily generate isolated code snippets rather than complete, large-scale project code. Without coding expertise, these users struggle to interpret, modify, and iteratively refine the outputs of LLMs, making it impossible to assemble a complete project. To address this issue, we propose Self-Rectified Large-Scale Code Generator (SRLCG), a framework that generates complete multi-file project code from a single prompt. SRLCG employs a novel multidimensional chain-of-thought (CoT) and self-rectification to guide LLMs in generating correct and robust code files, then integrates them into a complete and coherent project using our proposed dynamic backtracking algorithm. Experimental results show that SRLCG generates code 15x longer than DeepSeek-V3, 16x longer than GPT-4, and at least 10x longer than other leading CoT-based baselines. Furthermore, they confirm its improved correctness, robustness, and performance compared to baselines in large-scale code generation.