arXiv subjects

Siqi Liu

Publications and source records attributed to Siqi Liu.

At least 19 recordsLinked to original sources

arXiv2026

TuringLLM: Efficiently Scaling Foundation Models Toward Physical AI

We present Turing-20B-A2B, a 20B-parameter Mixture-of-Experts language model that activates approximately 2B parameters per token, designed for long-context and latency-sensitive physical AI applications. The model adopts Quantile Routing in a dynamic top-k configuration, enabling token-adaptive expert allocation while maintaining balanced expert utilization and a controlled average compute budget. During deployment, we further apply capacity-constrained routing to prompt prefill for more regular and efficient expert execution, while retaining dropless routing during pretraining. Turing-20B-A2B also employs a hybrid attention architecture that combines Lightning Attention with a small number of full-attention layers for efficient long-context modeling. The model is pretrained with a progressive three-stage curriculum and extended to a native context length of 128K through continued pretraining, with further inference-time extension to 512K using YaRN. Despite its compact active-parameter budget, Turing-20B-A2B achieves, at the base-model stage, overall general capability exceeding Qwen3-8B Base and approaching Qwen3.5-9B Base, while maintaining strong long-context performance and favorable prefill-latency scaling. These results demonstrate an effective balance among model capability, long-context scalability, and practical inference efficiency.

cs.AI↗

arXiv2026

Small-World Communication Fabrics for Neuromorphic Multicore-SoCs

As neuromorphic systems scale beyond a single core, inter-core event communication can become a dominant contributor to memory footprint, latency, and energy consumption. Biological neural systems address a similar scaling challenge through small-world organization, combining dense local connectivity with sparse long-range projections. In this work, we compare two recent multicore neuromorphic systems implemented in the same 22-nm FDSOI technology and explicitly optimized for such connectivity. The first, NeoCorAl, uses an asynchronous packet-switched tree with hierarchical multicast, whereas the second, MOSAIC, employs an RRAM-based, circuit-switched two-dimensional mesh that performs routing in memory. We examine the resulting trade-offs in routing flexibility, hop count, memory requirements, multicast efficiency, and scalability. We further study how the relative efficiency of tree- and mesh-based routing depends on communication locality in spatially-embedded, random, and layered networks. Finally, we discuss routing-aware training as a means of jointly optimizing neural connectivity, task performance, and hardware mappability.

cs.ET↗

arXiv2026

A Multimodal Foundation Model for Longitudinal Patient Representation and Scalable Insight Generation in Oncology

Precision oncology necessitates a longitudinal model of patient state that captures cancer evolution and treatment over time, integrating multimodal observations. We introduce the oFM, a foundation model developed on a real-world oncology cohort of 1.67 million cancer patients that integrates clinical trajectories with DNA, RNA, and H&E pathology. Patient-level partitions were reserved for training, validation, and testing, with over one million patients used for training. The oFM encodes daily clinical and molecular episodes and, along with pathology images, integrates them over time to produce a patient state embedding. We evaluate frozen oFM embeddings against expert-curated clinical and molecular baseline features. In prognostic benchmarks, the oFM improved AUC for treatment response, progression-free survival, and overall survival (0.774 vs. 0.563 for overall survival). Across 11 comparative-treatment cohorts, the oFM embeddings achieved a three-fold higher pooled and scale-normalized treatment-benefit AUTOC than baseline features with improved benefit ranking in 9 of 11 cohorts, and provided stronger prognostic discrimination within both treatment arms. We also evaluated a mechanism discovery framework that interprets downstream models built on oFM embeddings by linking their predicted outcomes to clinically and biologically grounded mechanisms through an evidence-grounded temporal graph, enabling evaluation in clinical and drug-development applications.

cs.LG↗

arXiv2026

What Makes Graph Unified? Principles and Generative Sliding-Window Transformer for Graph Foundation Models

Graph Foundation Models (GFMs) have recently emerged as a promising paradigm for general-purpose graph learning, aiming to learn reusable knowledge that generalizes across diverse graph domains and downstream tasks, reducing the need for specific model development. Achieving this goal requires reconciling the substantial heterogeneity in node features, graph structures, and semantic information across domains. Among them, heterogeneous node features constitute a fundamental input-level barrier, as their dimensionality and semantics vary substantially across datasets. Existing studies typically project or map heterogeneous node features into a fixed-dimensional space, often implicitly equating dimensional uniformity with effective feature unification. Yet dimensional consistency alone does not ensure that the unified features preserve informative semantics and capture transferable patterns that can support cross-domain knowledge transfer. To bridge this conceptual gap, we distill four desiderata for cross-domain graph feature unification: formal uniformity, cross-domain transferability, information preservation, and backbone compatibility. Guided by these principles, we propose SliGFM, a graph foundation model built upon topology-aware sliding-window feature encoding and generative reconstruction. SliGFM orders feature dimensions by topological smoothness and scans the reordered features with a shared sliding-window feature encoder, transforming heterogeneous features into a common space of ordered fixed-dimensional feature tokens. This formulation enables a smoothness-aware transformer to capture transferable relational patterns among feature tokens within each node, while the generative reconstruction objective encourages preservation of the original feature information.

cs.LG↗

arXiv2026

Paradoxes of Game Theoretic Equilibria and Price of Anarchy

For decades, static solution concepts (Nash, Correlated, and Coarse Correlated Equilibria) and the Price of Anarchy (PoA) have formed the bedrock of algorithmic game theory, with no-regret learning proving fast convergence to such game-theoretic equilibria. We show that reducing multi-agent learning to static equilibrium and black-box regret analysis obscures underlying dynamic disequilibrium and game theoretic bounds. First, interior Nash equilibria lack $C^1$ vector field information, meaning agents cannot distinguish aligned from strictly opposing incentives. Inheriting this geometry, the worst-case pure Nash equilibria dictating robust PoA bounds manifest as topologically unstable strict saddles, and in canonical congestion games, as global repellers supported on almost everywhere strictly dominated strategies. Anchoring efficiency guarantees to these unstable states causes algebraic sensitivity; we prove that accommodating all strictly positive affine costs renders the PoA unbounded. Furthermore, projecting learning trajectories onto the discrete simplex of correlated play systematically accommodates non-rationalizable behavior. Evaluating dynamics via Coarse Correlated Equilibria or proximal refinements fails to preclude strictly dominated strategies. Moreover, optimal $O(1/T)$ swap-regret minimization does not preclude macroscopic turbulence, manifesting as chaotic limit sets even in minimal games. Finally, we examine the non-atomic limit of congestion games. Though considered highly stable with tight sub-linear $Θ(p/\ln p)$ PoA bounds (where $p$ is the polynomial degree), we prove that under discrete-time learning, the unique equilibrium destabilizes into Li-Yorke chaos and global attractors whose time-averaged inefficiency degrades exponentially as $2^p$. These results necessitate re-evaluating worst-case equilibrium frameworks for dynamically grounded metrics.

cs.GT↗

arXiv2026

Distill Where the Student Goes: Teacher-Regularized RL for English-Evidence Cross-Lingual RAG

Cross-lingual retrieval-augmented generation (RAG) is often deployed in an English-evidence regime, where users query in diverse languages but retrieved passages remain English. In this setting, generation can fail despite strong base models: English evidence induces language drift (English or code-switching outputs) and models use evidence unreliably when producing non-English answers. We attribute these failures to two post-training challenges: (i) errors are prefix-dependent, so fixed-trajectory supervision suffers from prefix mismatch; and (ii) sequence-level (partly discrete / judge-based) rewards yield noisy credit assignment and high-variance updates. We propose TR-RAG, a teacher-regularized RL recipe that couples reward optimization with on-policy distillation on student-visited prefixes. A compact student samples on-policy answers, while a stronger frozen teacher is queried only on those prefixes and provides a prefix-wise student-to-teacher reverse-KL anchor. We further introduce a reward decomposition for English-evidence multilingual generation, combining language consistency, character 3-gram recall, and an LLM-judge score for evidence-grounded correctness. Across three benchmarks (BioASQ-ENKB5, Hotpot-ENKB5, and naturally multilingual MKQA) and two backbones, TR-RAG improves the composite of language adherence and evidence-grounded correctness over strong baselines. Crucially, the teacher anchor acts as a safety net: on in-domain languages it prevents the large language-consistency collapses (up to ~27 percentage points) that reward-only RL can suffer by drifting below even the base model, while on distant out-of-distribution languages, where reward-only RL stalls at the base model's ceiling, it still improves evidence grounding; and on character 3-gram recall the compact student sometimes surpasses its 70B teacher.

cs.CL↗

arXiv2026

Looped World Models

Current world models face a fundamental tension: faithful long-horizon simulation demands deep computation, but deeper models are expensive to deploy and prone to compounding errors. We resolve this by introducing Looped World Models (LoopWM), which are the first looped architectures for world modelling. Our method iteratively refines latent environment states through a parameter-shared transformer block. This yield up to 100x parameter efficiency over conventional approaches with adaptive computation that automatically scales depth to match the complexity of each prediction step. Orthogonal to scaling model size and training data, LoopWM establishes iterative latent depth as a new scaling axis for world simulation, which might significantly push the community forward.

cs.LG↗

arXiv2026

Virchow: A Million-Slide Digital Pathology Foundation Model

The use of artificial intelligence to enable precision medicine and decision support systems through the analysis of pathology images has the potential to revolutionize the diagnosis and treatment of cancer. Such applications will depend on models' abilities to capture the diverse patterns observed in pathology images. To address this challenge, we present Virchow, a foundation model for computational pathology. Using self-supervised learning empowered by the DINOv2 algorithm, Virchow is a vision transformer model with 632 million parameters trained on 1.5 million hematoxylin and eosin stained whole slide images from diverse tissue and specimen types, which is orders of magnitude more data than previous works. The Virchow model enables the development of a pan-cancer detection system with 0.949 overall specimen-level AUC across 17 different cancer types, while also achieving 0.937 AUC on 7 rare cancer types. The Virchow model sets the state-of-the-art on the internal and external image tile level benchmarks and slide level biomarker prediction tasks. The gains in performance highlight the importance of training on massive pathology image datasets, suggesting scaling up the data and network architecture can improve the accuracy for many high-impact computational pathology applications where limited amounts of training data are available.

eess.IV↗

arXiv2026

A Method for Neutron-Gamma Pulse Shape Discrimination of CLYC Detector Based on a Gated Residual-Linear Attention Network

The discrimination of neutron and gamma pulse shapes is a key technology in fields such as nuclear safety monitoring and radiation assessment. An enhanced recursive gated cyclic residual-sparse linear attention network is developed on the CLYC detector experimental platform to overcome weak noise resistance, limited feature extraction and inferior real-time performance of conventional algorithms. The experimental dataset comprises 19,971 samples, which were pre-processed and stratified for model training and testing. Results indicate that the proposed algorithm achieves a quality factor of 2.2, with a classification accuracy of 98.7% and a recall rate of 99.4%. It achieves an accuracy of 95.1% under the 20 dB low signal-to-noise ratio condition, exhibiting excellent anti-noise ability.With around 2.8 million parameters, the model takes merely 0.05 ms to process a single pulse on GPU, satisfying real-time monitoring and embedded deployment demands.

physics.ins-det↗

arXiv2026

Compatibility and Accuracy Verification of CADmesh-Based Complex Geometry Modeling in Geant4

Geant4 Monte Carlo simulation relies on the Constructive Solid Geometry (CSG) method for complex geometric modeling. This method has low efficiency and a high application threshold. Importing triangular facet formats such as STL/OBJ via CADmesh is a promising alternative, but systematic evaluations of format compatibility, geometric accuracy, and physical simulation deviations are lacking. Construct open-source experimental environment based on Geant4 11.0, CADmesh 1.3.0 and FreeCAD 1.0. We design high and low precision gradient test cases using simple geometric bodies and complex engineering models, and systematically evaluate the import success rate, facet loss rate, volume error, and particle transport dose deviation for STL and OBJ formats.The results show a 100% import success rate for both formats; the volume error rate is <= 0.018% for high-precision models and <= 0.288% for low-precision models. The two formats share the same vertex facet data structure. This study designs a general adaptive interface. The interface reduces the number of parsing code lines by about 70% and maintains geometric accuracy.Furthermore, the tetrahedral mesh loading takes 3.1 times longer than tessellated solids, but the simulation time can be reduced from 15194.3 s to 77.28 s.

cs.GR↗

arXiv2026

Rethinking 1-bit Optimization Leveraging Pre-trained Large Language Models

1-bit LLM quantization offers significant advantages in reducing storage and computational costs. However, existing methods typically train 1-bit LLMs from scratch, failing to fully leverage pre-trained models. This results in high training costs and notable accuracy degradation. We identify that the large gap between full precision and 1-bit representations makes naive adaptation difficult. In this paper, we introduce a consistent progressive training for both forward and backward, smoothly converting the full-precision weights into the binarized ones. Additionally, we incorporate binary-aware initialization and dual-scaling compensation to reduce the difficulty of progressive training and improve the performance. Experimental results on LLMs of various sizes demonstrate that our method outperforms existing approaches. Our results show that high-performance 1-bit LLMs can be achieved using pre-trained models, eliminating the need for expensive training from scratch.

cs.CL↗

arXiv2026

CoCo-SAM3: Harnessing Concept Conflict in Open-Vocabulary Semantic Segmentation

SAM3 advances open-vocabulary semantic segmentation by introducing a prompt-driven mask generation paradigm. However, in multi-class open-vocabulary scenarios, masks generated independently from different category prompts lack a unified and inter-class comparable evidence scale, often resulting in overlapping coverage and unstable competition. Moreover, synonymous expressions of the same concept tend to activate inconsistent semantic and spatial evidence, leading to intra-class drift that exacerbates inter-class conflicts and compromises overall inference stability. To address these issues, we propose CoCo-SAM3 (Concept-Conflict SAM3), which explicitly decouples inference into intra-class enhancement and inter-class competition. Our method first aligns and aggregates evidence from synonymous prompts to strengthen concept consistency. It then performs inter-class competition on a unified comparable scale, enabling direct pixel-wise comparisons among all candidate classes. This mechanism stabilizes multi-class inference and effectively mitigates inter-class conflicts. Without requiring any additional training, CoCo-SAM3 achieves consistent improvements across eight open-vocabulary semantic segmentation benchmarks.

cs.CV↗

arXiv2026

FACTS: Table Summarization via Offline Template Generation with Agentic Workflows

Query-focused table summarization requires generating natural language summaries of tabular data conditioned on a user query, enabling users to access insights beyond fact retrieval. Existing approaches face key limitations: table-to-text models require costly fine-tuning and struggle with complex reasoning, prompt-based LLM methods suffer from token-limit and efficiency issues while exposing sensitive data, and prior agentic pipelines often rely on decomposition, planning, or manual templates that lack robustness and scalability. To mitigate these issues, we introduce an agentic workflow, FACTS, a Fast, Accurate, and Privacy-Compliant Table Summarization approach via Offline Template Generation. FACTS produces offline templates, consisting of SQL queries and Jinja2 templates, which can be rendered into natural language summaries and are reusable across multiple tables sharing the same schema. It enables fast summarization through reusable offline templates, accurate outputs with executable SQL queries, and privacy compliance by sending only table schemas to LLMs. Evaluations on widely-used benchmarks show that FACTS consistently outperforms baseline methods, establishing it as a practical solution for real-world query-focused table summarization. Our code is available at https://github.com/BorealisAI/FACTS.

cs.CL↗

arXiv2026

Video-Only ToM: Enhancing Theory of Mind in Multimodal Large Language Models

As large language models (LLMs) continue to advance, there is increasing interest in their ability to infer human mental states and demonstrate a human-like Theory of Mind (ToM). Most existing ToM evaluations, however, are centered on text-based inputs, while scenarios relying solely on visual information receive far less attention. This leaves a gap, since real-world human-AI interaction typically requires multimodal understanding. In addition, many current methods regard the model as a black box and rarely probe how its internal attention behaves in multiple-choice question answering (QA). The impact of LLM hallucinations on such tasks is also underexplored from an interpretability perspective. To address these issues, we introduce VisionToM, a vision-oriented intervention framework designed to strengthen task-aware reasoning. The core idea is to compute intervention vectors that align visual representations with the correct semantic targets, thereby steering the model's attention through different layers of visual features. This guidance reduces the model's reliance on spurious linguistic priors, leading to more reliable multimodal language model (MLLM) outputs and better QA performance. Experiments on the EgoToM benchmark-an egocentric, real-world video dataset for ToM with three multiple-choice QA settings-demonstrate that our method substantially improves the ToM abilities of MLLMs. Furthermore, results on an additional open-ended generation task show that VisionToM enables MLLMs to produce free-form explanations that more accurately capture agents' mental states, pushing machine-human collaboration toward greater alignment.

cs.CV↗

arXiv2026

FEASTS and MHONGOOSE: HI Column Density Distribution at $z=0$ for $N_\mathrm{HI}>10^{17.8}\, \mathrm{cm}^{-2}$

We present the first $z=0$ HI column density distribution function, $f(N_\mathrm{HI})$, extending down to $\log (N_\mathrm{HI}/\mathrm{cm}^{-2})=17.8$. This was derived from high-sensitivity 21-cm emission-line imaging at $\sim$1 kpc resolution. At high-column-densities (19.8$< \log (N_\mathrm{HI}/\mathrm{cm}^{-2}) <$21.3), our results align with earlier $z=0$ studies but benefit from 100 times greater sensitivity. Comparisons with $z\sim3$ quasar absorption-line studies reveal that $f(N_\mathrm{HI})$ at $z=0$ is systematically lower by 0.1-0.4 dex for $19.2< \log (N_\mathrm{HI}/\mathrm{cm}^{-2}) <21$. However, the distributions become comparable at $17.8< \log (N_\mathrm{HI}/\mathrm{cm}^{-2}) <19.2$, suggesting weak evolution in this regime. Extrapolating the length incidence ($\mathrm{d}N/\mathrm{d}X$) for $\log (N_\mathrm{HI}/\mathrm{cm}^{-2}) >17.5$ implies a covering fraction ($f_\mathrm{cov}$) of $\sim0.7$ within 1-kpc-scale HI-detected pixels at $z=0$. Notably, for $17.8< \log (N_\mathrm{HI}/\mathrm{cm}^{-2}) <20$, impact parameters at a given $N_\mathrm{HI}$ are significantly lower than previous $z\sim0$ absorption-line results and TNG50 simulation predictions. This discrepancy indicates challenges in identifying galaxy counterparts for absorbers and in recovering low-column-density HI within cosmological simulations. Finally, we derive a covering fraction of 0.006 for $\log (N_\mathrm{HI}/\mathrm{cm}^{-2}) >17.8$ gas within the virial radius around Milky-Way-like galaxies. These findings provide new constraints on the baryonic flows and gaseous dynamics governing galaxy evolution.

astro-ph.GA↗

arXiv2026

LEDA: Latent Semantic Distribution Alignment for Multi-domain Graph Pre-training

Recent advances in generic large models, such as GPT and DeepSeek, have motivated the introduction of universality to graph pre-training, aiming to learn rich and generalizable knowledge across diverse domains using graph representations to improve performance in various downstream applications. However, most existing methods face challenges in learning effective knowledge from generic graphs, primarily due to simplistic data alignment and limited training guidance. The issue of simplistic data alignment arises from the use of a straightforward unification for highly diverse graph data, which fails to align semantics and misleads pre-training models. The problem with limited training guidance lies in the arbitrary application of in-domain pre-training paradigms to cross-domain scenarios. While it is effective in enhancing discriminative representation in one data space, it struggles to capture effective knowledge from many graphs. To address these challenges, we propose a novel Latent sEmantic Distribution Alignment (LEDA) model for universal graph pre-training. Specifically, we first introduce a dimension projection unit to adaptively align diverse domain features into a shared semantic space with minimal information loss. Furthermore, we design a variational semantic inference module to obtain the shared latent distribution. The distribution is then adopted to guide the domain projection, aligning it with shared semantics across domains and ensuring cross-domain semantic learning. LEDA exhibits strong performance across a broad range of graphs and downstream tasks. Remarkably, in few-shot cross-domain settings, it significantly outperforms in-domain baselines and advanced universal pre-training models.

cs.LG↗

arXiv2026

Embedding-Based Context-Aware Reranker

Retrieval-Augmented Generation (RAG) systems rely on retrieving relevant evidence from a corpus to support downstream generation. The common practice of splitting a long document into multiple shorter passages enables finer-grained and targeted information retrieval. However, it also introduces challenges when a correct retrieval would require inference across passages, such as resolving coreference, disambiguating entities, and aggregating evidence scattered across multiple sources. Many state-of-the-art (SOTA) reranking methods, despite utilizing powerful large pretrained language models with potentially high inference costs, still neglect the aforementioned challenges. Therefore, we propose Embedding-Based Context-Aware Reranker (EBCAR), a lightweight reranking framework operating directly on embeddings of retrieved passages with enhanced cross-passage understandings through the structural information of the passages and a hybrid attention mechanism, which captures both high-level interactions across documents and low-level relationships within each document. We evaluate EBCAR against SOTA rerankers on the ConTEB benchmark, demonstrating its effectiveness for information retrieval requiring cross-passage inference and its advantages in both accuracy and efficiency.

cs.CL↗

arXiv2026

Mixed Magnification Aggregation for Generalizable Region-Level Representations in Computational Pathology

In recent years, a standard computational pathology workflow has emerged where whole slide images are cropped into tiles, these tiles are processed using a foundation model, and task-specific models are built using the resulting representations. At least 15 different foundation models have been proposed, and the vast majority are trained exclusively with tiles using the 20$\times$ magnification. However, it is well known that certain histologic features can only be discerned with larger context windows and requires a pathologist to zoom in and out when analyzing a whole slide image. Furthermore, creating 224$\times$224 pixel crops at 20$\times$ leads to a large number of tiles per slide, which can be gigapixel in size. To more accurately capture multi-resolution features and investigate the possibility of reducing the number of representations per slide, we propose a region-level mixing encoder. Our approach jointly fuses image tile representations of a mixed magnification foundation model using a masked embedding modeling pretraining step. We explore a design space for pretraining the proposed mixed-magnification region aggregators and evaluate our models on transfer to biomarker prediction tasks representing various cancer types. Results demonstrate cancer dependent improvements in predictive performance, highlighting the importance of spatial context and understanding.

cs.CV↗