arXiv subjects

Xinrong Zhang

Publications and source records attributed to Xinrong Zhang.

14 recordsLinked to original sources

arXiv2026

Proving olympiad geometry theorems on a superconducting quantum processor

Automated theorem proving seeks to use computational systems to prove or disprove mathematical and logical statements [1, 2]. It underpins a wide range of applications, and enhancing theorem-proving capabilities remains a central objective in artificial intelligence [3]. Although recent neuro-symbolic systems have achieved remarkable progress [4-7], their operation is ultimately constrained by classical computational architectures. Quantum computing [8], by contrast, enables information encoding and coherent parallelism beyond classical limits [9-14], raising the possibility of accelerating structured symbolic deduction [15]. Here we report the experimental realization of automated geometry theorem proving on a fully programmable superconducting quantum processor. We develop two complementary quantum proving frameworks. The first implements Wu's algebraic elimination method using quantum pseudo-division, with multivariate polynomials represented in superposition states, enabling quantum algebraic theorem proving. The second implements the full-angle method as backward symbolic reasoning through a hybrid quantum strategy-guided architecture, demonstrating a general route toward quantum symbolic proof search. As illustrative examples, we prove two theorems on a superconducting quantum processor: the perpendicularity of the diagonals of a square and a 1978 International Mathematical Olympiad geometry problem. Our results establish, at the experimental level, automated logical reasoning as a viable task for near-term quantum processors and provide a concrete pathway toward quantum-enhanced symbolic intelligence.

quant-ph↗

arXiv2026

S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?

Large language models (LLMs) increasingly interact with external environments and accumulate substantial behavioral experience, yet existing agent benchmarks largely evaluate them as fixed policies. It therefore remains unclear whether an agent can actively test its behavior, judge the resulting experience, and use that experience to improve future decisions. We introduce \textbf{S\textsuperscript{3}Gym}, an interactive benchmark for evaluating LLM self-improvement through three coupled capabilities: \textbf{Self-Testing}, \textbf{Self-Judging}, and \textbf{Self-Improvement}. S$^3$Gym separates permissive exploration from strict held-out evaluation and instantiates this protocol in seven text-based games with executable environment verifiers. We evaluate three pathways for incorporating interaction experience: direct History ICL, score-conditioned Summary Memory, and parameter Training. Our experiments reveal that self-improvement is neither automatic nor uniform. Context-level experience improves performance for several model--game pairs, but the most effective pathway depends strongly on the task structure: summaries are beneficial when experience can be compressed into reusable strategic rules, yet often underperform raw history when success depends on precise, state-contingent information. Parameter training produces substantial gains on some tasks, but also exhibits unstable improvement and severe negative transfer on others. These findings show that recognizing successful actions is insufficient; agents must also transform feedback into executable and transferable policies. S$^3$Gym provides a unified framework for diagnosing this process and identifying the bottlenecks that prevent agents from translating interaction experience into reliable self-improvement.

cs.CL↗

arXiv2026

A superconducting surface-code processor with lattice-surgery logical operations

Fault-tolerant logical operations are fundamental for scalable quantum computation. Here, we report the experimental realization of lattice-surgery operations between a pair of distance-three surface-code logical qubits on a planar superconducting processor. During repeated syndrome extraction cycles, the logical qubits exhibit per-cycle error rates of $0.0365(2)$ and $0.0282(1)$, respectively, after leakage events are rejected. By leveraging joint initialization and lattice splitting, we deterministically prepare a logical Bell state, confirming genuine bipartite entanglement via the error-corrected logical state fidelity. We further execute a two-qubit Deutsch-Jozsa algorithm at the logical level to demonstrate algorithmic utility in a fault-tolerant framework. Finally, to achieve universal control, we implement magic-state injection and gate teleportation to realize continuous non-Clifford rotations about the logical $X$ axis. For the logical $R_{X}(π/4)$ gate, we achieve a logical gate fidelity of $0.943_{-9}^{+10}$ conditioned on the absence of detected errors. These results establish lattice surgery as a practical and versatile paradigm for logical computation in near-term surface-code architectures, representing a critical milestone toward scalable fault-tolerant quantum advantage in superconducting circuits.

quant-ph↗

arXiv2026

Teleportation transition of surface codes on a superconducting quantum processor

The topological surface code is a leading candidate for harnessing long-range entanglement to protect logical quantum information against errors, and teleportation of logical states is desirable for robust quantum information processing. Nevertheless, scaling up the surface code in quantum teleportation poses a formidable challenge to experiment. Here on a superconducting quantum processor with 125 qubits, we demonstrate the robust teleportation of topological rotated surface code prepared by a linear-depth unitary circuit, with code distances up to 7. We obtain the teleportation phase diagram by tuning the local entangling gates uniformly across a finite threshold. Furthermore, we show that the entangling threshold can be boosted by coherent qubit rotations that inject magic resources beyond the Clifford regime, restoring the duality symmetry of the topological phase, which serves as a guiding principle to minimize the entanglement resource. Our results shed light on simulating and leveraging topological quantum matter on quantum devices, and pave the way to the ultimate goal of distributed fault tolerant quantum computation.

quant-ph↗

arXiv2026

Beyond the Turn-Based Game: Enabling Real-Time Conversations with Duplex Models

As large language models (LLMs) increasingly permeate daily lives, there is a growing demand for real-time interactions that mirror human conversations. Traditional turn-based chat systems driven by LLMs prevent users from verbally interacting with the system while it is generating responses. To overcome these limitations, we adapt existing LLMs to \textit{duplex models} so that these LLMs can listen for users while generating output and dynamically adjust themselves to provide users with instant feedback. % such as in response to interruptions. Specifically, we divide the queries and responses of conversations into several time slices and then adopt a time-division-multiplexing (TDM) encoding-decoding strategy to pseudo-simultaneously process these slices. Furthermore, to make LLMs proficient enough to handle real-time conversations, we build a fine-tuning dataset consisting of alternating time slices of queries and responses as well as covering typical feedback types in instantaneous interactions. Our experiments show that although the queries and responses of conversations are segmented into incomplete slices for processing, LLMs can preserve their original performance on standard benchmarks with a few fine-tuning steps on our dataset. Automatic and human evaluation indicate that duplex models make user-AI interactions more natural and human-like, and greatly improve user satisfaction compared to vanilla LLMs. Our duplex model and dataset will be released.

cs.CL↗

arXiv2026

Stuffed Mamba: Oversized States Lead to the Inability to Forget

Recent advancements in recurrent architectures, such as Mamba and RWKV, have showcased strong language capabilities. Unlike transformer-based models, these architectures encode all contextual information into a fixed-size state, leading to great inference efficiency. However, this approach can cause information interference, where different token data conflicts, resulting in performance degradation and incoherent outputs beyond a certain context length. To prevent this, most RNNs incorporate mechanisms designed to "forget" earlier tokens. In this paper, we reveal that Mamba-based models struggle to effectively forget earlier tokens even with built-in forgetting mechanisms. We demonstrate that this issue stems from training on contexts that are too short for the state size, enabling the model to perform well without needing to learn how to forget. Then, we show that the minimum training length required for the model to learn forgetting scales linearly with the state size, and the maximum context length for accurate retrieval of a 5-digit passkey scales exponentially with the state size, indicating that the model retains some information beyond the point where forgetting begins. These findings highlight a critical limitation in current RNN architectures and provide valuable insights for improving long-context modeling. Our work suggests that future RNN designs must account for the interplay between state size, training length, and forgetting mechanisms to achieve robust performance in long-context tasks.

cs.CL↗

arXiv2025

Mining the Social Fabric: Unveiling Communities for Fake News Detection in Short Videos

Short video platforms have become a major medium for information sharing, but their rapid content generation and algorithmic amplification also enable the widespread dissemination of fake news. Detecting misinformation in short videos is challenging due to their multi-modal nature and the limited context of individual videos. While recent methods focus on analyzing content signals-visual, textual, and audio-they often overlook implicit relationships among videos, uploaders, and events. To address this gap, we propose DugFND (Dual-community graph for fake news detection), a novel method that enhances existing video classifiers by modeling two key community patterns: (1) uploader communities, where uploaders with shared interests or similar content creation patterns group together, and (2) event-driven communities, where videos related to the same or semantically similar public events form localized clusters. We construct a heterogeneous graph connecting uploader, video, and event nodes, and design a time-aware heterogeneous graph attention network to enable effective message passing. A reconstruction-based pretraining phase further improves node representation learning. DugFND can be applied to any pre-trained classifier. Experiments on public datasets show that our method achieves significant performance gains, demonstrating the value of dual-community modeling for fake news detection in short videos.

cs.MM↗

arXiv2024

Seq1F1B: Efficient Sequence-Level Pipeline Parallelism for Large Language Model Training

The emergence of large language models (LLMs) relies heavily on distributed training strategies, among which pipeline parallelism plays a crucial role. As LLMs' training sequence length extends to 32k or even 128k, the current pipeline parallel methods face severe bottlenecks, including high memory footprints and substantial pipeline bubbles, greatly hindering model scalability and training throughput. To enhance memory efficiency and training throughput, in this work, we introduce an efficient sequence-level one-forward-one-backward (1F1B) pipeline scheduling method tailored for training LLMs on long sequences named Seq1F1B. Seq1F1B decomposes batch-level schedulable units into finer sequence-level units, reducing bubble size and memory footprint. Considering that Seq1F1B may produce slight extra bubbles if sequences are split evenly, we design a computation-wise strategy to partition input sequences and mitigate this side effect. Compared to competitive pipeline baseline methods such as Megatron 1F1B pipeline parallelism, our method achieves higher training throughput with less memory footprint. Notably, Seq1F1B efficiently trains a LLM with 30B parameters on sequences up to 64k using 64 NVIDIA A100 GPUs without recomputation strategies, a feat unachievable with existing methods. Our source code is based on Megatron-LM, and now is avaiable at: https://github.com/MayDomine/Seq1F1B.git.

cs.DC↗

arXiv2024

Ouroboros: Generating Longer Drafts Phrase by Phrase for Faster Speculative Decoding

Speculative decoding is a widely used method that accelerates the generation process of large language models (LLMs) with no compromise in model performance. It achieves this goal by using an existing smaller model for drafting and then employing the target LLM to verify the draft in a low-cost parallel manner. Under such a drafting-verification framework, drafting efficiency has become a bottleneck in the final speedup of speculative decoding. Therefore, generating longer drafts at less cost can lead to better decoding speedup. To achieve this, we introduce Ouroboros, which can generate draft phrases to parallelize the drafting process and meanwhile lengthen drafts in a training-free manner. The experimental results on various typical text generation tasks show that Ouroboros can achieve speedups of up to $2.8\times$ over speculative decoding and $3.9\times$ over vanilla decoding, without fine-tuning draft and target models. The source code of Ouroboros is available at https://github.com/thunlp/Ouroboros.

cs.CL↗

arXiv2024

MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategies

The burgeoning interest in developing Large Language Models (LLMs) with up to trillion parameters has been met with concerns regarding resource efficiency and practical expense, particularly given the immense cost of experimentation. This scenario underscores the importance of exploring the potential of Small Language Models (SLMs) as a resource-efficient alternative. In this context, we introduce MiniCPM, specifically the 1.2B and 2.4B non-embedding parameter variants, not only excel in their respective categories but also demonstrate capabilities on par with 7B-13B LLMs. While focusing on SLMs, our approach exhibits scalability in both model and data dimensions for future LLM research. Regarding model scaling, we employ extensive model wind tunnel experiments for stable and optimal scaling. For data scaling, we introduce a Warmup-Stable-Decay (WSD) learning rate scheduler (LRS), conducive to continuous training and domain adaptation. We present an in-depth analysis of the intriguing training dynamics that occurred in the WSD LRS. With WSD LRS, we are now able to efficiently study data-model scaling law without extensive retraining experiments on both axes of model and data, from which we derive the much higher compute optimal data-model ratio than Chinchilla Optimal. Additionally, we introduce MiniCPM family, including MiniCPM-DPO, MiniCPM-MoE and MiniCPM-128K, whose excellent performance further cementing MiniCPM's foundation in diverse SLM applications. MiniCPM models are available publicly at https://github.com/OpenBMB/MiniCPM .

cs.CL↗

arXiv2024

Predicting Emergent Abilities with Infinite Resolution Evaluation

The scientific scale-up of large language models (LLMs) necessitates a comprehensive understanding of their scaling properties. However, the existing literature on the scaling properties only yields an incomplete answer: optimization loss decreases predictably as the model size increases, in line with established scaling law; yet no scaling law for task has been established and the task performances are far from predictable during scaling. Task performances typically show minor gains on small models until they improve dramatically once models exceed a size threshold, exemplifying the ``emergent abilities''. In this study, we discover that small models, although they exhibit minor performance, demonstrate critical and consistent task performance improvements that are not captured by conventional evaluation strategies due to insufficient measurement resolution. To measure such improvements, we introduce PassUntil, an evaluation strategy with theoretically infinite resolution, through massive sampling in the decoding phase. With PassUntil, we conduct a quantitative investigation into the scaling law of task performance. The investigation contains two parts. Firstly, a strict task scaling law that is not conventionally known to exist, is identified, enhancing the predictability of task performances. Remarkably, we are able to predict the performance of the 2.4B model on code generation with merely 0.05\% deviation before training starts, which is the first systematic attempt to verify predictable scaling proposed by GPT-4's report. Secondly, we are able to study emergent abilities quantitatively. We identify a kind of accelerated emergence whose scaling curve cannot be fitted by standard scaling law function and has a increasing speed. We then examine two hypothesis and imply that the ``multiple circuits hypothesis'' might be responsible for the accelerated emergence.

cs.CL↗

arXiv2024

$\infty$Bench: Extending Long Context Evaluation Beyond 100K Tokens

Processing and reasoning over long contexts is crucial for many practical applications of Large Language Models (LLMs), such as document comprehension and agent construction. Despite recent strides in making LLMs process contexts with more than 100K tokens, there is currently a lack of a standardized benchmark to evaluate this long-context capability. Existing public benchmarks typically focus on contexts around 10K tokens, limiting the assessment and comparison of LLMs in processing longer contexts. In this paper, we propose $\infty$Bench, the first LLM benchmark featuring an average data length surpassing 100K tokens. $\infty$Bench comprises synthetic and realistic tasks spanning diverse domains, presented in both English and Chinese. The tasks in $\infty$Bench are designed to require well understanding of long dependencies in contexts, and make simply retrieving a limited number of passages from contexts not sufficient for these tasks. In our experiments, based on $\infty$Bench, we evaluate the state-of-the-art proprietary and open-source LLMs tailored for processing long contexts. The results indicate that existing long context LLMs still require significant advancements to effectively process 100K+ context. We further present three intriguing analyses regarding the behavior of LLMs processing long context.

cs.CL↗

arXiv2022

Phocus: Picking Valuable Research from a Sea of Citations

The deluge of new papers has significantly blocked the development of academics, which is mainly caused by author-level and publication-level evaluation metrics that only focus on quantity. Those metrics have resulted in several severe problems that trouble scholars focusing on the important research direction for a long time and even promote an impetuous academic atmosphere. To solve those problems, we propose Phocus, a novel academic evaluation mechanism for authors and papers. Phocus analyzes the sentence containing a citation and its contexts to predict the sentiment towards the corresponding reference. Combining others factors, Phocus classifies citations coarsely, ranks all references within a paper, and utilizes the results of the classifier and the ranking model to get the local influential factor of a reference to the citing paper. The global influential factor of the reference to the citing paper is the product of the local influential factor and the total influential factor of the citing paper. Consequently, an author's academic influential factor is the sum of his contributions to each paper he co-authors.

cs.DL↗

arXiv2021

Estimate Metabolite Taxonomy and Structure with a Fragment-Centered Database and Fragment Network

Metabolite structure identification has become the major bottleneck of the mass spectrometry based metabolomics research. Till now, number of mass spectra databases and search algorithms have been developed to address this issue. However, two critical problems still exist: the low chemical component record coverage in databases and significant MS/MS spectra variations related to experiment equipment and parameter settings. In this work, we considered the molecule fragment as basic building blocks of the metabolic components which had relatively consistent signatures in MS/MS spectra. And from a bottom-up point of view, we built a fragment centered database, MSFragDB, by reorganizing the data from the Human Metabolome Database (HMDB) and developed an intensity-free searching algorithm to search and rank the most relative metabolite according to the users' input. We also proposed the concept of fragment network, a graph structure that encoded the relationship between the molecule fragments to find close motif that indicated a specific chemical structure. Although based on the same dataset as the HMDB, validation results implied that the MSFragDB had a higher hit ratio and furthermore, estimated possible taxonomy that a query spectrum belongs to when the corresponding chemical component was missing in the database. Aid by the Fragment Network, the MSFragDB was also proved to be able to estimate the right structure while the MS/MS spectrum suffers from the precursor-contamination. The strategy proposed is general and can be adopted in existing databases. We believe MSFragDB and Fragment Network can improve the performance of structure identification with existing data. The beta version of the database is freely available at www.xrzhanglab.com/msfragdb/.

q-bio.QM↗