Hyperparameter optimization (HPO) plays a critical role in improving model performance. Transformer-based HPO methods have shown great potential; however, existing approaches rely heavily on large-scale historical optimization trajectories and lack effective reinforcement learning (RL) techniques, thereby limiting their efficiency and performance improvements. Inspired by the success of Group Relative Policy Optimization (GRPO) in large language models (LLMs), we propose GRPOformer -- a novel hyperparameter optimization framework that integrates reinforcement learning (RL) with Transformers. In GRPOformer, Transformers are employed to generate new hyperparameter configurations from historical optimization trajectories, while GRPO enables rapid trajectory construction and optimization strategy learning from scratch. Moreover, we introduce Policy Churn Regularization (PCR) to enhance the stability of GRPO training. Experimental results on OpenML demonstrate that GRPOformer consistently outperforms baseline methods across diverse tasks, offering new insights into the application of RL for HPO.

Abstract--Large Language Models are transforming software engineering, yet prompt management in practice remains ad hoc, hindering reliability, reuse, and integration into industrial workflows. We present Prompt-with-Me, a practical solution for structured prompt management embedded directly in the development environment. The system automatically classifies prompts using a four-dimensional taxonomy encompassing intent, author role, software development lifecycle stage, and prompt type. T o enhance prompt reuse and quality, Prompt-with-Me suggests language refinements, masks sensitive information, and extracts reusable templates from a developer's prompt library. Our taxonomy study of 1,108 real-world prompts demonstrates that modern LLMs can accurately classify software engineering prompts. Furthermore, our user study with 11 participants shows strong developer acceptance, with high usability (Mean SUS=73), low cognitive load (Mean NASA-TLX=21), and reported gains in prompt quality and efficiency through reduced repetitive effort. Lastly, we offer actionable insights for building the next generation of prompt management and maintenance tools for software engineering workflows. Large Language Models (LLMs) such as GPT-4 and Claude are rapidly transforming modern software engineering (SE). Tools powered by these models have moved from experimental prototypes to integral components of everyday development workflows. For instance, industry reports on GitHub Copilot, a GPT-powered AI pair programmer, show that developers using the tool complete coding tasks up to 55% faster, while 85% report increased confidence in their code quality [1]. These numbers highlight not only the efficiency gains brought by LLMs but also their growing influence on developer experience and software quality. Despite the transformative role of LLMs in software engineering, the primary interface for human-LLM interaction, the prompt, remains surprisingly informal. Prompts are typically written ad hoc, often with fragmented grammar or inconsistent wording, even though minor phrasing changes can significantly alter model behavior [2]. Unlike source code, prompts are rarely versioned, reviewed, or maintained across the software development lifecycle, leaving critical interactions with LLMs effectively unmanaged.

We introduce FreqRank, a mutation-based defense to localize malicious components in LLM outputs and their corresponding backdoor triggers. FreqRank assumes that the malicious sub-string(s) consistently appear in outputs for triggered inputs and uses a frequency-based ranking system to identify them. Our ranking system then leverages this knowledge to localize the backdoor triggers present in the inputs. We create nine malicious models through fine-tuning or custom instructions for three downstream tasks, namely, code completion (CC), code generation (CG), and code summarization (CS), and show that they have an average attack success rate (ASR) of 86.6%. Furthermore, FreqRank's ranking system highlights the malicious outputs as one of the top five suggestions in 98% of cases. We also demonstrate that FreqRank's effectiveness scales as the number of mutants increases and show that FreqRank is capable of localizing the backdoor trigger effectively even with a limited number of triggered samples. Finally, we show that our approach is 35-50% more effective than other defense methods.

Next point-of-interest (POI) recommendation predicts a user's next destination from historical movements. Traditional models require intensive training, while LLMs offer flexible and generalizable zero-shot solutions but often generate generic or geographically irrelevant results due to missing trajectory and spatial context. To address these issues, we propose RALLM-POI, a framework that couples LLMs with retrieval-augmented generation and self-rectification. We first propose a Historical Trajectory Retriever (HTR) that retrieves relevant past trajectories to serve as contextual references, which are then reranked by a Geographical Distance Reranker (GDR) for prioritizing spatially relevant trajectories. Lastly, an Agentic LLM Rectifier (ALR) is designed to refine outputs through self-reflection. Without additional training, RALLM-POI achieves substantial accuracy gains across three real-world Foursquare datasets, outperforming both conventional and LLM-based baselines.

Recently, deep learning has driven significant advancements in multivariate time series forecasting (MTSF) tasks. However, much of the current research in MTSF tends to evaluate models from a holistic perspective, which obscures the individual contributions and leaves critical issues unaddressed. Adhering to the current modeling paradigms, this work bridges these gaps by systematically decomposing deep MTSF methods into their core, fine-grained components like series-patching tokenization, channel-independent strategy, attention modules, or even Large Language Models and Time-series Foundation Models. Through extensive experiments and component-level analysis, our work offers more profound insights than previous benchmarks that typically discuss models as a whole. Furthermore, we propose a novel automated solution called TSGym for MTSF tasks. Unlike traditional hyperparameter tuning, neural architecture searching or fixed model selection, TSGym performs fine-grained component selection and automated model construction, which enables the creation of more effective solutions tailored to diverse time series data, therefore enhancing model transferability across different data sources and robustness against distribution shifts. Extensive experiments indicate that TSGym significantly outperforms existing state-of-the-art MTSF and AutoML methods. All code is publicly available on https://github.com/SUFE-AILAB/TSGym.

Contact-rich manipulation is crucial for robots to perform tasks requiring precise force control, such as insertion, assembly, and in-hand manipulation. However, most imitation learning (IL) policies remain position-centric and lack explicit force awareness, and adding force/torque sensors to collaborative robot arms is often costly and requires additional hardware design. To overcome these issues, we propose FILIC, a Force-guided Imitation Learning framework with impedance torque control. FILIC integrates a Transformer-based IL policy with an impedance controller in a dual-loop structure, enabling compliant force-informed, force-executed manipulation. For robots without force/torque sensors, we introduce a cost-effective end-effector force estimator using joint torque measurements through analytical Jacobian-based inversion while compensating with model-predicted torques from a digital twin. We also design complementary force feedback frameworks via handheld haptics and VR visualization to improve demonstration quality. Experiments show that FILIC significantly outperforms vision-only and joint-torque-based methods, achieving safer, more compliant, and adaptable contact-rich manipulation. Our code can be found in https://github.com/TATP-233/FILIC.

Recent studies have suggested a processing framework for multilingual inputs in decoder-based LLMs: early layers convert inputs into English-centric and language-agnostic representations; middle layers perform reasoning within an English-centric latent space; and final layers generate outputs by transforming these representations back into language-specific latent spaces. However, the internal dynamics of such transformation and the underlying mechanism remain underexplored. Towards a deeper understanding of this framework, we propose and empirically validate The Transfer Neurons Hypothesis: certain neurons in the MLP module are responsible for transferring representations between language-specific latent spaces and a shared semantic latent space. Furthermore, we show that one function of language-specific neurons, as identified in recent studies, is to facilitate movement between latent spaces. Finally, we show that transfer neurons are critical for reasoning in multilingual LLMs.

Large Reasoning Models (LRMs) often suffer from computational inefficiency due to overthinking, where a fixed reasoning budget fails to match the varying complexity of tasks. To address this issue, we propose Adaptive Overclocking, a method that makes the overclocking hyperparameter $α$ dynamic and context-aware. Our method adjusts reasoning speed in real time through two complementary signals: (1) token-level model uncertainty for fine-grained step-wise control, and (2) input complexity estimation for informed initialization. We implement this approach with three strategies: Uncertainty-Aware Alpha Scheduling (UA-$α$S), Complexity-Guided Alpha Initialization (CG-$α$I), and a Hybrid Adaptive Control (HAC) that combines both. Experiments on GSM8K, MATH, and SVAMP show that HAC achieves superior accuracy-latency trade-offs, reducing unnecessary computation on simple problems while allocating more resources to challenging ones. By mitigating overthinking, Adaptive Overclocking enhances both efficiency and overall reasoning performance.

As robots proliferate in manufacturing, Design for Robotic Assembly (DfRA), which is designing products for efficient automated assembly, is increasingly important. Traditional approaches to DfRA rely on manual planning, which is time-consuming, expensive and potentially impractical for complex objects. Large language models (LLM) have exhibited proficiency in semantic interpretation and robotic task planning, stimulating interest in their application to the automation of DfRA. But existing methodologies typically rely on heuristic strategies and rigid, hard-coded physics simulators that may not translate into real-world assembly contexts. In this work, we present Iterative Design for Robotic Assembly (IDfRA), a framework using iterative cycles of planning, execution, verification, and re-planning, each informed by self-assessment, to progressively enhance design quality within a fixed yet initially under-specified environment, thereby eliminating the physics simulation with the real world itself. The framework accepts as input a target structure together with a partial environmental representation. Through successive refinement, it converges toward solutions that reconcile semantic fidelity with physical feasibility. Empirical evaluation demonstrates that IDfRA attains 73.3\% top-1 accuracy in semantic recognisability, surpassing the baseline on this metric. Moreover, the resulting assembly plans exhibit robust physical feasibility, achieving an overall 86.9\% construction success rate, with design quality improving across iterations, albeit not always monotonically. Pairwise human evaluation further corroborates the advantages of IDfRA relative to alternative approaches. By integrating self-verification with context-aware adaptation, the framework evidences strong potential for deployment in unstructured manufacturing scenarios.

Direct Preference Optimization (DPO) is a powerful paradigm to align language models with human preferences using pairwise comparisons. However, its binary win-or-loss supervision often proves insufficient for training small models with limited capacity. Prior works attempt to distill information from large teacher models using behavior cloning or KL divergence. These methods often focus on mimicking current behavior and overlook distilling reward modeling. To address this issue, we propose \textit{Teacher Value-based Knowledge Distillation} (TVKD), which introduces an auxiliary reward from the value function of the teacher model to provide a soft guide. This auxiliary reward is formulated to satisfy potential-based reward shaping, ensuring that the global reward structure and optimal policy of DPO are preserved. TVKD can be integrated into the standard DPO training framework and does not require additional rollouts. Our experimental results show that TVKD consistently improves performance across various benchmarks and model sizes.