Yingyu Lin

Jiongli Zhu, Haoquan Guan, Parjanya Prajakta Prashant et al.

Text-to-SQL ensembles improve over single-candidate generation by drawing multiple SQL candidates and selecting one, but their effectiveness is bounded by Pass@K, the probability that at least one of K candidates is correct. Existing methods source diversity heuristically through stochastic decoding or prompt variants, leaving candidate sets dominated by correlated failures. We present DivSkill-SQL, a residual skill optimization framework that builds complementary agentic Text-to-SQL ensembles without model fine-tuning: each new skill is optimized on examples the current skill ensemble fails on, provably targeting its marginal contribution to Pass@K. On Spider2-Lite, DivSkill-SQL improves selected accuracy by up to +11.1 points on Snowflake and +8.3 on BigQuery over the strongest ensemble baseline, with consistent gains across two base models (Opus-4.6 and GPT-5.4). Skills optimized on a single dialect transfer without retraining across dialects (Snowflake, BigQuery, SQLite) and to a different task formulation, such as BIRD-Critic (+2.6 pts). Error diagnostics show up to 3x fewer hallucinated schema references and function calls, indicating that gains come from genuinely reliable complementary skills rather than surface-form variation.

Yingyu Lin, Yi-An Ma, Yu-Xiang Wang et al.

Posterior sampling, i.e., exponential mechanism to sample from the posterior distribution, provides $\varepsilon$-pure differential privacy (DP) guarantees and does not suffer from potentially unbounded privacy breach introduced by $(\varepsilon,δ)$-approximate DP. In practice, however, one needs to apply approximate sampling methods such as Markov chain Monte Carlo (MCMC), thus re-introducing the unappealing $δ$-approximation error into the privacy guarantees. To bridge this gap, we propose the Approximate SAample Perturbation (abbr. ASAP) algorithm which perturbs an MCMC sample with noise proportional to its Wasserstein-infinity ($W_\infty$) distance from a reference distribution that satisfies pure DP or pure Gaussian DP (i.e., $δ=0$). We then leverage a Metropolis-Hastings algorithm to generate the sample and prove that the algorithm converges in $W_\infty$ distance. We show that by combining our new techniques with a localization step, we obtain the first nearly linear-time algorithm that achieves the optimal rates in the DP-ERM problem with strongly convex and smooth losses.

Xunpeng Huang, Yingyu Lin, Nikki Lijing Kuang et al.

Continuous diffusion models have demonstrated remarkable performance in data generation across various domains, yet their efficiency remains constrained by two critical limitations: (1) the local adjacency structure of the forward Markov process, which restricts long-range transitions in the data space, and (2) inherent biases introduced during the simulation of time-inhomogeneous reverse denoising processes. To address these challenges, we propose Quantized Transition Diffusion (QTD), a novel approach that integrates data quantization with discrete diffusion dynamics. Our method first transforms the continuous data distribution $p_*$ into a discrete one $q_*$ via histogram approximation and binary encoding, enabling efficient representation in a structured discrete latent space. We then design a continuous-time Markov chain (CTMC) with Hamming distance-based transitions as the forward process, which inherently supports long-range movements in the original data space. For reverse-time sampling, we introduce a \textit{truncated uniformization} technique to simulate the reverse CTMC, which can provably provide unbiased generation from $q_*$ under minimal score assumptions. Through a novel KL dynamic analysis of the reverse CTMC, we prove that QTD can generate samples with $O(d\ln^2(d/ε))$ score evaluations in expectation to approximate the $d$--dimensional target distribution $p_*$ within an $ε$ error tolerance. Our method not only establishes state-of-the-art inference efficiency but also advances the theoretical foundations of diffusion-based generative modeling by unifying discrete and continuous diffusion paradigms.

Yingyu Lin, Erchi Wang, Yi-An Ma et al.

We propose a framework to convert $(\varepsilon, δ)$-approximate Differential Privacy (DP) mechanisms into $(\varepsilon', 0)$-pure DP mechanisms under certain conditions, a process we call ``purification.'' This algorithmic technique leverages randomized post-processing with calibrated noise to eliminate the $δ$ parameter while achieving near-optimal privacy-utility tradeoff for pure DP. It enables a new design strategy for pure DP algorithms: first run an approximate DP algorithm with certain conditions, and then purify. This approach allows one to leverage techniques such as strong composition and propose-test-release that require $δ>0$ in designing pure-DP methods with $δ=0$. We apply this framework in various settings, including Differentially Private Empirical Risk Minimization (DP-ERM), stability-based release, and query release tasks. To the best of our knowledge, this is the first work with a statistically and computationally efficient reduction from approximate DP to pure DP. Finally, we illustrate the use of this reduction for proving lower bounds under approximate DP constraints with explicit dependence in $δ$, avoiding the sophisticated fingerprinting code construction.

Haoran Deng, Yingyu Lin, Zhenghao Lin et al.

Long-context language models unlock advanced capabilities in reasoning, code generation, and document summarization by leveraging dependencies across extended spans of text. However, a significant portion of readily available long-text data lacks meaningful long-distance dependencies; most spans can be predicted using only local context. Training on such data is inefficient, making careful data selection crucial. Therefore, we introduce LongFilter, a framework for curating training data tailored to long-context pretraining. LongFilter measures the information gain provided by extended context by contrasting model predictions under long-context versus short-context settings, thereby identifying samples where long-range dependencies are essential. Experiments with LLaMA-3-8B, extending its context length from 8K to 64K, show that LongFilter efficiently selects high-quality data and yields substantial improvements on benchmarks such as HELMET, LongBench, and RULER.

Xunpeng Huang, Yingyu Lin, Nishant Jain et al.

We study masked discrete diffusion -- a flexible paradigm for text generation in which tokens are progressively corrupted by special mask symbols before being denoised. Although this approach has demonstrated strong empirical performance, its theoretical complexity in high-dimensional settings remains insufficiently understood. Existing analyses largely focus on uniform discrete diffusion, and more recent attempts addressing masked diffusion either (1) overlook widely used Euler samplers, (2) impose restrictive bounded-score assumptions, or (3) fail to showcase the advantages of masked discrete diffusion over its uniform counterpart. To address this gap, we show that Euler samplers can achieve $ε$-accuracy in total variation (TV) with $\tilde{O}(d^{2}ε^{-3/2})$ discrete score evaluations, thereby providing the first rigorous analysis of typical Euler sampler in masked discrete diffusion. We then propose a Mask-Aware Truncated Uniformization (MATU) approach that both removes bounded-score assumptions and preserves unbiased discrete score approximation. By exploiting the property that each token can be unmasked at most once, MATU attains a nearly $ε$-free complexity of $O(d\,\ln d\cdot (1-ε^2))$. This result surpasses existing uniformization methods under uniform discrete diffusion, eliminating the $\ln(1/ε)$ factor and substantially speeding up convergence. Our findings not only provide a rigorous theoretical foundation for masked discrete diffusion, showcasing its practical advantages over uniform diffusion for text generation, but also pave the way for future efforts to analyze diffusion-based language models developed under masking paradigm.

Huiwei Zhou, Zhuang Liu, Shixian Ning et al.

Automatically extracting Protein-Protein Interactions (PPI) from biomedical literature provides additional support for precision medicine efforts. This paper proposes a novel memory network-based model (MNM) for PPI extraction, which leverages prior knowledge about protein-protein pairs with memory networks. The proposed MNM captures important context clues related to knowledge representations learned from knowledge bases. Both entity embeddings and relation embeddings of prior knowledge are effective in improving the PPI extraction model, leading to a new state-of-the-art performance on the BioCreative VI PPI dataset. The paper also shows that multiple computational layers over an external memory are superior to long short-term memory networks with the local memories.

Huiwei Zhou, Zhuang Liu1, Shixian Ning et al.

Protein-protein interaction (PPI) extraction from published scientific literature provides additional support for precision medicine efforts. However, many of the current PPI extraction methods need extensive feature engineering and cannot make full use of the prior knowledge in knowledge bases (KB). KBs contain huge amounts of structured information about entities and relationships, therefore plays a pivotal role in PPI extraction. This paper proposes a knowledge-aware attention network (KAN) to fuse prior knowledge about protein-protein pairs and context information for PPI extraction. The proposed model first adopts a diagonal-disabled multi-head attention mechanism to encode context sequence along with knowledge representations learned from KB. Then a novel multi-dimensional attention mechanism is used to select the features that can best describe the encoded context. Experiment results on the BioCreative VI PPI dataset show that the proposed approach could acquire knowledge-aware dependencies between different words in a sequence and lead to a new state-of-the-art performance.

Huiwei Zhou, Shixian Ning, Yunlong Yang et al.

Chemical-disease relation (CDR) extraction is significantly important to various areas of biomedical research and health care. Nowadays, many large-scale biomedical knowledge bases (KBs) containing triples about entity pairs and their relations have been built. KBs are important resources for biomedical relation extraction. However, previous research pays little attention to prior knowledge. In addition, the dependency tree contains important syntactic and semantic information, which helps to improve relation extraction. So how to effectively use it is also worth studying. In this paper, we propose a novel convolutional attention network (CAN) for CDR extraction. Firstly, we extract the shortest dependency path (SDP) between chemical and disease pairs in a sentence, which includes a sequence of words, dependency directions, and dependency relation tags. Then the convolution operations are performed on the SDP to produce deep semantic dependency features. After that, an attention mechanism is employed to learn the importance/weight of each semantic dependency vector related to knowledge representations learned from KBs. Finally, in order to combine dependency information and prior knowledge, the concatenation of weighted semantic dependency representations and knowledge representations is fed to the softmax layer for classification. Experiments on the BioCreative V CDR dataset show that our method achieves comparable performance with the state-of-the-art systems, and both dependency information and prior knowledge play important roles in CDR extraction task.

Huiwei Zhou, Yunlong Yang, Shixian Ning et al.

Automatically extracting the relationships between chemicals and diseases is significantly important to various areas of biomedical research and health care. Biomedical experts have built many large-scale knowledge bases (KBs) to advance the development of biomedical research. KBs contain huge amounts of structured information about entities and relationships, therefore plays a pivotal role in chemical-disease relation (CDR) extraction. However, previous researches pay less attention to the prior knowledge existing in KBs. This paper proposes a neural network-based attention model (NAM) for CDR extraction, which makes full use of context information in documents and prior knowledge in KBs. For a pair of entities in a document, an attention mechanism is employed to select important context words with respect to the relation representations learned from KBs. Experiments on the BioCreative V CDR dataset show that combining context and knowledge representations through the attention mechanism, could significantly improve the CDR extraction performance while achieve comparable results with state-of-the-art systems.

Huiwei Zhou, Chengkun Lang, Zhuang Liu et al.

Background: Automatic extraction of chemical-disease relations (CDR) from unstructured text is of essential importance for disease treatment and drug development. Meanwhile, biomedical experts have built many highly-structured knowledge bases (KBs), which contain prior knowledge about chemicals and diseases. Prior knowledge provides strong support for CDR extraction. How to make full use of it is worth studying. Results: This paper proposes a novel model called "Knowledge-guided Convolutional Networks (KCN)" to leverage prior knowledge for CDR extraction. The proposed model first learns knowledge representations including entity embeddings and relation embeddings from KBs. Then, entity embeddings are used to control the propagation of context features towards a chemical-disease pair with gated convolutions. After that, relation embeddings are employed to further capture the weighted context features by a shared attention pooling. Finally, the weighted context features containing additional knowledge information are used for CDR extraction. Experiments on the BioCreative V CDR dataset show that the proposed KCN achieves 71.28% F1-score, which outperforms most of the state-of-the-art systems. Conclusions: This paper proposes a novel CDR extraction model KCN to make full use of prior knowledge. Experimental results demonstrate that KCN could effectively integrate prior knowledge and contexts for the performance improvement.