Improving the plot progression ability of role-playing language agents (RPLAs) plays a critical role in enhancing user experience and dialogue immersion. However, research on plot progression in role-playing language agents is still limited. Existing datasets often contain too few dialogue turns and lack large-scale sentences with clear deterministic plot turns, which restricts further model development and evaluation. To bridge this gap, we propose RoleArena, an innovative multi-agent role-playing environment. By introducing a critic agent and combining discrete plot lines with environment-agent-based dynamic judgments, the system generates long multi-turn dialogues with key markers for plot progression. Based on RoleArena, we construct the PlotStream dataset, which contains 39.3k characters and 170k dialogues, with annotations for key plot progression sentences. In addition, based on PlotStream, we further develop PlotRole-7B and PlotRole-72B, RPLAs built upon the Qwen2.5 model and specialized in plot progression capabilities. Extensive experiments show that RoleArena and PlotRole improve dialogue generation with plot progression and strengthen the ability of plot progression.

Large Language Models have shown promise in translating natural language into executable optimization models, yet they often suffer from the Sisyphus Dilemma: a memoryless cycle where identical errors are repeated across structurally similar problems. Existing retrieval-augmented strategies primarily fetch static problem-model pairs as few-shot demonstrators, failing to capture the dynamic reasoning required to resolve execution failures. To bridge this gap, we propose EOM, a framework that implements Experience Replay to transform transient rectification steps into persistent knowledge. EOM distills interaction histories into Causal Correction Mappings, indexing both diagnostic insights and prohibitive traps. By utilizing a structure-aware retrieval mechanism that aligns semantic intent with abstract syntax trees and solver tracebacks, the system enables models to recall specific correction strategies for isomorphic errors. Extensive experiments across seven benchmarks demonstrate that EOM improves modeling accuracy by 8.45% on complex tasks while reducing token consumption by 28.65% and interaction turns by 25.82%, validating the efficiency of a “Rectify Once, Solve Many” paradigm.

Large language models (LLMs) and vision-language models (VLMs) are increasingly used as optimization assistants to produce solutions, generate solver-executable programs, or both. However, current evaluations are misaligned with deployment in three ways: they (P1) fail to represent multimodal problem specifications, (P2) score outcomes only and cannot localize where failures occur along the modeling pipeline, and (P3) rarely report inference cost, obscuring reliability–cost trade-offs. We introduce Graph Optimization benchmark (GOBench), an aligned multimodal benchmark with solver-derived oracles and a four-layer diagnostic protocol that evaluates intermediate artifacts as well as end results, together with the Visual Inference Penalty (VIP) to measure multimodal overhead. Across frontier and open-weight models under paired text-only vs. T+V settings, we find that vision reliably increases inference cost, while its reliability impact is regime-dependent: frontier models often benefit from visual grounding, whereas several mid-tier/open models exhibit a Visual Paradox where vision reduces downstream executability and verification coverage. End-to-end success is frequently bottlenecked by intermediate-stage dropout; supervised fine-tuning on intermediate targets can mitigate this attrition in open models, enabling a reproducible harness for diagnosing failure modes and quantifying reliability–cost trade-offs.

Mixture-of-Experts (MoE) models offer a promising path for scaling model capacity, yet their massive memory footprint poses significant challenges for deployment on resource-constrained edge devices. Existing solutions, such as static pruning or dynamic offloading, often struggle to balance model accuracy with inference latency due to irreversible information loss or prohibitive I/O overhead. In this paper, we propose LightMoE, a novel framework for memory-efficient MoE inference that exploits the inherent functional redundancy and temporal locality of expert activation. LightMoE employs a frequency-aware expert initialization strategy to retain a compact core of resident experts and introduces a similarity-based redirection mechanism to compensate for missing experts without incurring I/O costs. Furthermore, it incorporates a lightweight runtime manager that performs coarse-grained, task-level expert replacement to adapt to shifting data distributions. Empirical evaluations on representative edge platforms demonstrate that LightMoE achieves a superior accuracy-efficiency trade-off, improving average accuracy by 4.3% over static pruning and 2.4% over dynamic swapping methods, while maintaining inference latency comparable to strictly pruned models.

Large language models make natural-language-to-MILP modeling increasingly plausible, but reliably selecting among multiple generated programs remains challenging: structurally similar errors can create false majorities, while solving every candidate integer program is costly. We propose Structural-Semantic Consistency (SSC), a best-of-K selector combining structural consistency, static quality, text alignment, and two diagnostics obtained from continuous LP relaxations. Notably, SSC does not require solving any candidate integer programs during selection. Across seven benchmarks and 618 fixed candidate pools, SSC obtains 55.7% accuracy, compared with a random baseline of 51.5%, 53.7% for a zero-shot listwise LLM judge, and 55.3% for a leave-one-benchmark-out learned ranker. The symbolic fusion reaches 53.6%, with LP diagnostics adding 2.1 percentage points. These results establish multi-signal weak verification as a competitive and interpretable baseline for candidate selection.

Multi-hop question answering requires retrieving and integrating multiple pieces of evidence distributed across documents. Recent graph-based methods improve multi-hop reasoning through graph traversal over relational evidence structures. However, existing methods rely on independent path ranking during evidence organization, which often produces redundant and reasoning-incomplete evidence sets due to substantial structural overlap among selected graph paths. In this work, we argue that graph-based evidence organization may, to some extent, be formulated as a constrained evidence subgraph optimization problem. Based on this perspective, we propose CROG, a coverage-aware reasoning optimization framework that performs marginal-gain-based path selection to favor previously uncovered reasoning steps while suppressing redundant evidence, with node-budget and connectivity constraints to encourage compact and coherent evidence subgraphs. Experiments on multiple benchmarks demonstrate that CROG achieves superior reasoning performance while using fewer evidence units on average, and analysis based on exhaustive optimal search further validates the effectiveness of the greedy optimization strategy.

Pre-trained language models (PLMs) are widely used in NLP but struggle with capturing entity knowledge. To address this, knowledge enhancement techniques have been proposed. However, existing methods rely heavily on external knowledge bases embedding and often introduce noisy entity representations. In this work, we propose a novel Knowledge Enhancement Filtering Framework named KEFF, which contains both knowledge enhancement and knowledge enhancement filtering modules for PLM. We find that there are certain redundant bits in the embedding space of PLMs. Building on this insight, we implement knowledge-enhanced mapping of redundant bit values in entity span tokens. In order to solve the knowledge enhancement problem of existing methods that introduce noisy entity representation knowledge, we further propose a novel knowledge enhancement filter based on our knowledge enhancement method. Finally, experiments on four knowledge-driven NLP tasks show that our method effectively improves the ability of PLMs on downstream tasks. Compared to state-of-the-art approachs, our method achieves the highest F1-score and accuracy, while reducing the computational cost by 1.7-2.5x.

Lyric-to-melody generation is an important task in automatic songwriting. Previous lyric-to-melody generation systems usually adopt end-to-end models that directly generate melodies from lyrics, which suffer from several issues: 1) lack of paired lyric-melody training data; 2) lack of control on generated melodies. In this paper, we develop TeleMelody, a two-stage lyric-to-melody generation system with music template (e.g., tonality, chord progression, rhythm pattern, and cadence) to bridge the gap between lyrics and melodies (i.e., the system consists of a lyric-to-template module and a template-to-melody module). TeleMelody has two advantages. First, it is data efficient. The template-to-melody module is trained in a self-supervised way (i.e., the source template is extracted from the target melody) that does not need any lyric-melody paired data. The lyric-to-template module is made up of some rules and a lyric-to-rhythm model, which is trained with paired lyric-rhythm data that is easier to obtain than paired lyric-melody data. Second, it is controllable. The design of the template ensures that the generated melodies can be controlled by adjusting the musical elements in the template. Both subjective and objective experimental evaluations demonstrate that TeleMelody generates melodies with higher quality, better controllability, and less requirement on paired lyric-melody data than previous generation systems.

Error correction is widely used in automatic speech recognition (ASR) to post-process the generated sentence, and can further reduce the word error rate (WER). Although multiple candidates are generated by an ASR system through beam search, current error correction approaches can only correct one sentence at a time, failing to leverage the voting effect from multiple candidates to better detect and correct error tokens. In this work, we propose FastCorrect 2, an error correction model that takes multiple ASR candidates as input for better correction accuracy. FastCorrect 2 adopts non-autoregressive generation for fast inference, which consists of an encoder that processes multiple source sentences and a decoder that generates the target sentence in parallel from the adjusted source sentence, where the adjustment is based on the predicted duration of each source token. However, there are some issues when handling multiple source sentences. First, it is non-trivial to leverage the voting effect from multiple source sentences since they usually vary in length. Thus, we propose a novel alignment algorithm to maximize the degree of token alignment among multiple sentences in terms of token and pronunciation similarity. Second, the decoder can only take one adjusted source sentence as input, while there are multiple source sentences. Thus, we develop a candidate predictor to detect the most suitable candidate for the decoder. Experiments on our inhouse dataset and AISHELL-1 show that FastCorrect 2 can further reduce the WER over the previous correction model with single candidate by 3.2% and 2.6%, demonstrating the effectiveness of leveraging multiple candidates in ASR error correction. FastCorrect 2 achieves better performance than the cascaded re-scoring and correction pipeline and can serve as a unified post-processing module for ASR.

Unsupervised neural machine translation (NMT) has attracted a lot of attention recently. While state-of-the-art methods for unsupervised translation usually perform well between similar languages (e.g., English-German translation), they perform poorly between distant languages, because unsupervised alignment does not work well for distant languages. In this work, we introduce unsupervised pivot translation for distant languages, which translates a language to a distant language through multiple hops, and the unsupervised translation on each hop is relatively easier than the original direct translation. We propose a learning to route (LTR) method to choose the translation path between the source and target languages. LTR is trained on language pairs whose best translation path is available and is applied on the unseen language pairs for path selection. Experiments on 20 languages and 294 distant language pairs demonstrate the advantages of the unsupervised pivot translation for distant languages, as well as the effectiveness of the proposed LTR for path selection. Specifically, in the best case, LTR achieves an improvement of 5.58 BLEU points over the conventional direct unsupervised method.