Retrieval-augmented generation (RAG) has proven to be effective in mitigating hallucinations in large language models, yet its effectiveness remains limited in complex, multi-step reasoning scenarios.Recent efforts have incorporated search-based interactions into RAG, enabling iterative reasoning with real-time retrieval. Most approaches rely on outcome-based supervision, offering no explicit guidance for intermediate steps. This often leads to reward hacking and degraded response quality. We propose Bi-RAR, a novel retrieval-augmented reasoning framework that evaluates each intermediate step jointly in both forward and backward directions. To assess the information completeness of each step, we introduce a bidirectional information distance grounded in Kolmogorov complexity, approximated via language model generation probabilities. This quantification measures both how far the current reasoning is from the answer and how well it addresses the question. To optimize reasoning under these bidirectional signals, we adopt a multi-objective reinforcement learning framework with a cascading reward structure that emphasizes early trajectory alignment. Empirical results on seven question answering benchmarks demonstrate that Bi-RAR surpasses previous methods and enables efficient interaction and reasoning with the search engine during training and inference.

Recent LLM-based TTS systems achieve strong quality and zero-shot ability, but lack fine-grained emotional control due to their reliance on discrete speech tokens. Existing approaches either limit emotions to categorical labels or cannot generalize to LLM-based architectures. We propose EMORL-TTS (Fine-grained Emotion-controllable TTS with Reinforcement Learning), a framework that unifies global intensity control in the VAD space with local emphasis regulation. Our method combines supervised fine-tuning with reinforcement learning guided by task-specific rewards for emotion category, intensity, and emphasis. Moreover, we further investigate how emphasis placement modulates fine-grained emotion intensity. Experiments show that EMORL-TTS improves emotion accuracy, intensity differentiation, and emphasis clarity, while preserving synthesis quality comparable to strong LLM-based baselines. Synthesized samples are available on-line1.

Temporal misalignment arising from heterogeneous hardware sampling rates and network jitter poses a critical challenge in embodied intelligence, especially for consumer-grade domestic robots, creating a fundamental mismatch between the policy’s perceived state and the physical world. Standard approaches typically rely on implicit alignment or rigid synchronization, which fail to balance temporal fidelity with information completeness. In this work, we introduce Time-Consistent Synchronization and Completion (TCSC), a unified protocol designed to mitigate this mismatch without altering policy architectures. We first formalize temporal inconsistency using a Wasserstein-based metric and analyze its impact on joint information structure. Our analysis reveals a non-monotonic trade-off: while tightening history windows improves temporal consistency, excessive strictness degrades performance due to modality dropout. TCSC resolves this by enforcing strict causal synchronization constraints during inference while leveraging offline interpolation-based completion during training to align data distributions end-to-end. Extensive evaluations across simulated and real-world domestic tasks (e.g., smart home object tidy-up on Cobot Magic) demonstrate that TCSC significantly improves policy robustness against asynchronous perturbations, outperforming baselines that neglect the temporal structure of multimodal data.

Large language models fine-tuned on instruction–code pairs may memorize and subsequently leak sensitive training data. Existing differentially private (DP) code generation methods primarily protect code snippets while assuming prompts are public, which fails in realistic scenarios where prompts may also contain sensitive information. When prompts cannot be explicitly learned or used during generation, code synthesis suffers from severe utility degradation and reduced diversity. To address these challenges, we propose PrivCode++, the first work to explore DP code generation under where both prompts and code snippets are considered sensitive in LLM fine-tuning. PrivCode++ introduces a two-stage DP framework with a Privacy-Free Latent Conditioning module, enabling effective DP fine-tuning and data synthesis without direct access to sensitive prompts or code. Extensive experiments show that PrivCode++ achieves substantially higher utility than baselines, remains competitive with the method with relaxing privacy assumptions, and provides stronger privacy guarantees.

Recent advances in neural information retrieval models have significantly enhanced these models’ effectiveness across information retrieval tasks. The robustness of these models, which is essential for ensuring their reliability in practice, has also garnered significant attention. With a wide array of research on robust information retrieval being published, we believe it is the opportune moment to consolidate the current status, glean insights from existing methodologies, and lay the groundwork for future development. Robustness of information retrieval is a multifaceted concept and we emphasize the importance of robustness against performance variance, out-of-distribution scenarios, and adversarial attacks. With a focus on out-of-distribution and adversarial robustness, we dissect robustness solutions for dense retrieval models and neural ranking models, respectively, recognizing them as pivotal components of the neural information retrieval pipeline. We provide an in-depth discussion of methods, datasets, and evaluation metrics, shedding light on challenges and future directions in the era of large language models. To accompany this survey, we release three additional resources: (1) a curated list of publications related to robust information retrieval, (2) a tutorial based on this survey, and (3) a heterogeneous benchmark for robust information retrieval, BestIR, that collects all known datasets for evaluating information retrieval systems for robustness. We hope that this study provides useful clues for future research on the robustness of information retrieval models and helps to develop trustworthy IR systems.

This article explores an event-triggered self-learning parallel tracking control strategy for continuous-time nonlinear systems with actuator faults. To mitigate the impact of actuator faults, an enhanced performance index function is developed based on the unknown upper bound of such faults. Furthermore, an augmented error nonlinear system is constructed by incorporating the tracking error and control input of the original system, so as to implement parallel tracking control via the adaptive dynamic programming (ADP) approach. By designing an adaptive law, the unknown upper bound of faults is estimated online, which adjusts the control parameters in real-time and compensates for the loss of control effectiveness induced by faults. A critic network is employed to approximate the optimal value function within an event-triggered framework, while a dynamic triggering mechanism is adopted to reduce communication overhead. Through the design of appropriate Lyapunov functions, the asymptotic stability of the closed-loop system is rigorously proven, and the weight estimation error is demonstrated to be uniformly ultimately bounded (UUB). Finally, the simulation results are provided to validate the effectiveness of the proposed control strategy.

Answering real-world open-domain multi-hop questions over massive corpora is a critical challenge in Retrieval-Augmented Generation (RAG) systems. Recent research employs reinforcement learning (RL) to end-to-end optimize the retrieval-augmented reasoning process, directly enhancing its capacity to resolve complex queries. However, reliable deployment is hindered by two obstacles. 1) Retrieval Collapse: iterative retrieval over large corpora fails to locate intermediate evidence containing bridge answers without reasoning-guided planning, causing downstream reasoning to collapse. 2) Learning Instability: end-to-end trajectory training suffers from weak credit assignment across reasoning chains and poor error localization across modules, causing overfitting to benchmark-specific heuristics that limit transferability and stability. To address these problems, we propose PRISMA, a decoupled RL-guided framework featuring a Plan-Retrieve-Inspect-Solve-Memoize architecture. PRISMA's strength lies in reasoning-guided collaboration: the Inspector provides reasoning-based feedback to refine the Planner's decomposition and fine-grained retrieval, while enforcing evidence-grounded reasoning in the Solver. We optimize individual agent capabilities via Two-Stage Group Relative Policy Optimization (GRPO). Stage I calibrates the Planner and Solver as specialized experts in planning and reasoning, while Stage II utilizes Observation-Aware Residual Policy Optimization (OARPO) to enhance the Inspector's ability to verify context and trigger targeted recovery. Experiments show that PRISMA achieves state-of-the-art performance on ten benchmarks and can be deployed efficiently in real-world scenarios.

Retrieval-augmented generation (RAG) enhances large language model (LLM) reasoning by retrieving external documents, but also opens up new attack surfaces. We study knowledge-base poisoning attacks in RAG, where an attacker injects malicious content into the retrieval corpus, which is then naturally surfaced by the retriever and consumed by the LLM during reasoning. Unlike prior work that floods the corpus with poisoned documents, we propose AdversarialCoT, a query-specific attack that poisons only a single document in the corpus. AdversarialCoT first extracts the target LLM's reasoning framework to guide the construction of an initial adversarial chain-of-thought (CoT). The adversarial document is iteratively refined through interactions with the LLM, progressively exposing and exploiting critical reasoning vulnerabilities. Experiments on benchmark LLMs show that a single adversarial document can significantly degrade reasoning accuracy, revealing subtle yet impactful weaknesses. This study exposes security risks in RAG systems and provides actionable insights for designing more robust LLM reasoning pipelines.

Retrieval-augmented generation (RAG) can achieve strong answer accuracy on multi-hop questions, but outcome-level rewards often leave reasoning traces weakly grounded and difficult to audit. Under noisy retrieval, models may exhibit right-answer-wrong-reason failures, where the final answer is correct but the supporting rationale exploits shortcuts or unsupported evidence. We therefore ask whether faithfulness-guided alignment hurts answer accuracy in post-retrieval reasoning. To study this question, we propose CRAFT (Calibrated Reasoning with Answer-Faithful Traces), a reinforcement learning framework for the response-generation stage of retrieval-augmented multi-hop question answering. CRAFT trains models to produce structured reasoning traces with configurable auditability, while combining deterministic rewards for format compliance, answer correctness, and citation validity with a judge-based reward for semantic faithfulness. Experiments across model scales and benchmarks show that CRAFT unlocks task-specific reasoning capacity from 1.5B upward, improving both answer accuracy and Faithfulness; at 0.5B, performance remains sharply template-dependent. At 7B, CRAFT improves Faithfulness over the Base model in all evaluated settings and remains competitive with strong closed-source models. Code is available at https://github.com/Ameame1/CRAFT.

Retrieval-augmented generation (RAG) has emerged as a promising paradigm for mitigating hallucinations in large language models (LLMs).However, the intrinsic heterogeneity between the retriever and the generator often leads to a mismatch between retrieved evidence and generation needs, hindering effective coordination.We argue that competition between discriminative retrieval and generative modeling can more effectively expose their mutual weaknesses and induce deeper interaction. Motivated by this insight, we propose CARL (Co-opetition AdveRsarial Learning), a framework that formulates retriever–generator training in RAG as a minimax game. In this game, the retriever is optimized to retrieve both useful and adversarially useless documents to challenge the generator, while the generator learns to identify useful evidence and remain robust to misleading retrievals to produce accurate answers.Experiments on seven benchmark datasets demonstrate that CARL consistently improves RAG performance, validating the effectiveness of adversarial co-opetition in enhancing retriever–generator synergy.

Efficient motion planning with the error tolerance is crucial for dynamic robotic tasks, particularly robotic table tennis. This task demands simultaneous high efficiency and error tolerance. First, the incoming ball's high speed allows only tens of milliseconds for motion planning. Second, two types of errors, namely, ball-paddle motion uncertainty and joint limit violation, must be tolerated to ensure a high success rate of planning (SRP) and striking. This article proposes an advanced joint planning framework designed for high efficiency and error tolerance. To tolerate the error of ball-paddle motion uncertainty, this work introduces a joint classification criterion according to the joint motion characteristics. To tolerate the error of joint limit violation, based on the classification criterion, this study also develops a robust reference trajectory generation scheme, named error-tolerant-variable-sigmoid-based motion template (ETVSMT), to fully consider the motion capabilities of different joints. The ETVSMT approach utilizes the variable-sigmoid-based motion template (VSMT) as the backbone and designs its basic and remedial portions to tolerate the error of hard joint limit violation. The implementation of the ETVSMT scheme results in an average SRP of 98% and a ball-striking success rate of up to 90%, with execution time on an Intel Xeon CPU as low as 15 ms. Furthermore, the proposed method ensures the outgoing ball lands on the opposite side of the table with an average landing error of 19.62 cm and a net-passing height error of 12.68 cm. The proposed method can also benefit other dynamic tasks like human-robot interaction to improve the planning efficiency.

We explore adversarial attacks against retrieval-augmented generation (RAG) systems to identify their vulnerabilities. We focus on generating human-imperceptible adversarial examples and introduce a novel imperceptible retrieve-to-generate attack against RAG. This task aims to find imperceptible perturbations that retrieve a target document, originally excluded from the initial top-k candidate set, in order to influence the final answer generation. To address this task, we propose ReGENT, a reinforcement learning-based framework that tracks interactions between the attacker and the target RAG and continuously refines attack strategies based on relevance-generation-naturalness rewards. Experiments on newly constructed factual and non-factual question-answering benchmarks demonstrate that ReGENT significantly outperforms existing attack methods in misleading RAG systems with small imperceptible text perturbations.

Realizing energy-distributed cooperative scheduling while guaranteeing user privacy is challenging for community microgrid energy management. A dual-interaction deep deterministic policy gradient (DI-DDPG) method based on the contribution mechanism is proposed in this paper to realize distributed co-scheduling of energy in community microgrids under user privacy, electricity demand, and uncertainty of renewable energy sources and loads. To avoid privacy risks, each residential microgrid is defined as an agent with learning capabilities, and model training and decision-making are executed locally. The contribution mechanism encourages users to participate in co-scheduling in the community through contribution evaluation and benefit distribution. Meanwhile, to cope with the uncertainty of renewable energy and loads, a mid-day scheduling algorithm is introduced to adjust the day-ahead scheduling strategy and combine it with the demand response mechanism, greatly improving the model’s performance under uncertainty. Finally, simulation experiments using open-source real data and periodic data are conducted, and the results verified the effectiveness of the contribution mechanism and the mid-day scheduling algorithm in DI-DDPG and its excellent performance in energy co-scheduling.

Recent advances in generative modeling enable neural networks to generate weights without relying on gradient-based optimization. However, current methods are limited by issues of over-coupling and long-horizon. The former tightly binds weight generation with task-specific objectives, thereby limiting the flexibility of the learned optimizer. The latter leads to inefficiency and low accuracy during inference, caused by the lack of local constraints. In this paper, we propose Lo-Hp, a decoupled two-stage weight generation framework that enhances flexibility through learning various optimization policies. It adopts a hybrid-policy sub-trajectory balance objective, which integrates on-policy and off-policy learning to capture local optimization policies. Theoretically, we demonstrate that learning solely local optimization policies can address the long-horizon issue while enhancing the generation of global optimal weights. In addition, we validate Lo-Hp's superior accuracy and inference efficiency in tasks that require frequent weight updates, such as transfer learning, few-shot learning, domain generalization, and large language model adaptation.

Memory capability is a critical aspect of large language models (LLMs). However, the disparity in memory ability between small and large LLMs remains unclear. In this paper, we present a novel investigation into the memory capabilities of both small and large LLMs, introducing an innovative knowledge-based dataset, enriched with knowledge frequency annotations. Derived from Wikidata, this dataset quantifies knowledge frequency by counting the co-occurrences of head and tail entities of knowledge graph triples in part of the pre-training corpus. Building upon this, we constructed a fact-based question-answering dataset called KDF, and evaluated the memory performance of state-of-the-art pre-trained base model families. Our comprehensive experiments demonstrate that large LLMs exhibit robust memory capabilities, retaining most facts even when they occur infrequently. Conversely, small LLMs are limited to recalling only a subset of high-frequency facts, struggling significantly with low-frequency information. Our study not only illuminates the memory discrepancies between different scales of LLMs but also offers a valuable resource and methodology for future research in LLMs.

Product search is a crucial component of modern e-commerce platforms, with billions of user queries every day. In product search systems, first-stage retrieval should achieve high recall while ensuring efficient online deployment. Sparse retrieval is particularly attractive in this context due to its interpretability and storage efficiency. However, sparse retrieval methods suffer from severe vocabulary mismatch issues, leading to suboptimal performance in product search scenarios.With their potential for semantic analysis, large language models (LLMs) offer a promising avenue for mitigating vocabulary mismatch issues and thereby improving retrieval quality. Directly applying LLMs to sparse retrieval in product search exposes two key challenges:(1)Queries and product titles are typically short and highly susceptible to LLM-induced hallucinations, such as generating irrelevant expansion terms or underweighting critical literal terms like brand names and model numbers;(2)The large vocabulary space of LLMs leads to difficulty in initializing training effectively, making it challenging to learn meaningful sparse representations in such ultra-high-dimensional spaces.To address these challenges, we propose PROSPER, a framework for PROduct search leveraging LLMs as SParsE Retrievers. PROSPER incorporates: (1)A literal residual network that alleviates hallucination in lexical expansion by reinforcing underweighted literal terms through a residual compensation mechanism; and (2)A lexical focusing window that facilitates effective training initialization via a coarse-to-fine sparsification strategy.Extensive offline and online experiments show that PROSPER significantly outperforms sparse baselines and achieves recall performance comparable to advanced dense retrievers, while also achieving revenue increments online.

Large language models have revolutionized sign language generation by automatically transforming text into high-quality sign language videos, providing accessible communication for the Deaf community. However, existing LLM-based approaches prioritize semantic accuracy while overlooking emotional expressions, resulting in outputs that lack naturalness and expressiveness. We propose EASL (Emotion-Aware Sign Language), a multi-emotion-guided generation architecture for fine-grained emotional integration. We introduce emotion-semantic disentanglement modules with progressive training to separately extract semantic and affective features. During pose decoding, the emotional representations guide semantic interaction to generate sign poses with 7-class emotion confidence scores, enabling emotional expression recognition. Experimental results demonstrate that EASL achieves pose accuracy superior to all compared baselines by integrating multi-emotion information and effectively adapts to diffusion models to generate expressive sign language videos.

Existing GAN inversion methods work brilliantly in reconstructing high-quality (HQ) images while struggling with more common low-quality (LQ) inputs in practical application. To address this issue, we propose Unsupervised Domain Adaptation (UDA) in the inversion process, namely UDA-inversion, for effective inversion and editing of both HQ and LQ images. Regarding unpaired HQ images as the source domain and LQ images as the unlabeled target domain, we introduce a theoretical guarantee: loss value in the target domain is upper-bounded by loss in the source domain and a novel discrepancy function measuring the difference between two domains. Following that, we can only minimize this upper bound to obtain accurate latent codes for HQ and LQ images. Thus, constructive representations of HQ images can be spontaneously learned and transformed into LQ images without supervision. UDA-Inversion achieves a better PSNR of 22.14 on FFHQ dataset and performs comparably to supervised methods.