arXiv subjects

Zhijian Xu

Publications and source records attributed to Zhijian Xu.

At least 19 recordsLinked to original sources

arXiv2026

Fidel-TS: A High-Fidelity Multimodal Benchmark for Time Series Forecasting

The evaluation of time series forecasting models is hindered by a lack of high-quality benchmarks, leading to overestimated assessments of progress. Existing datasets suffer from issues ranging from small-scale, low-frequency, pre-training data contamination in unimodal designs to the temporal and description leakage prevalent in early multimodal designs. To address this, we formalize the core principles of high-fidelity benchmarking, focusing on data sourcing integrity, leak-free design, and structural clarity. We introduce Fidel-TS, a new large-scale benchmark built from these principles. Our experiments reveal the limitations of prior benchmarks and the potential discrepancies in model evaluation, providing new insights into multiple existing unimodal and multimodal forecasting models and LLMs across various evaluation tasks.

cs.LG↗

arXiv2026

Position: Universal Time Series Foundation Models Rest on a Category Error

This position paper argues that the pursuit of "Universal Foundation Models for Time Series" rests on a fundamental category error, mistaking a structural Container for a semantic Modality. We contend that because time series hold incompatible generative processes (e.g., finance vs. fluid dynamics), monolithic models degenerate into expensive "Generic Filters" that fail to generalize under distributional drift. To address this, we introduce the "Autoregressive Blindness Bound," a theoretical limit proving that history-only models cannot predict intervention-driven regime shifts. We advocate replacing universality with a Causal Control Agent paradigm, where an agent leverages external context to orchestrate a hierarchy of specialized solvers, from frozen domain experts to lightweight Just-in-Time adaptors. We conclude by calling for a shift in benchmarks from "Zero-Shot Accuracy" to "Drift Adaptation Speed" to prioritize robust, control-theoretic systems.

cs.AI↗

arXiv2026

From Samples to Scenarios: A New Paradigm for Probabilistic Forecasting

Most state-of-the-art probabilistic time series forecasting models rely on sampling to represent future uncertainty. However, this paradigm suffers from inherent limitations, such as lacking explicit probabilities, inadequate coverage, and high computational costs. In this work, we introduce \textbf{Probabilistic Scenarios}, an alternative paradigm designed to address the limitations of sampling. It operates by directly producing a finite set of \{Scenario, Probability\} pairs, thus avoiding Monte Carlo-like approximation. To validate this paradigm, we propose \textbf{TimePrism}, a simple model composed of only three parallel linear layers. Surprisingly, TimePrism achieves 9 out of 10 state-of-the-art results across five benchmark datasets on two metrics. The effectiveness of our paradigm comes from a fundamental reframing of the learning objective. Instead of modeling an entire continuous probability space, the model learns to represent a set of plausible scenarios and corresponding probabilities. Our work demonstrates the potential of the Probabilistic Scenarios paradigm, opening a promising research direction in forecasting beyond sampling.

cs.LG↗

arXiv2026

Making Slow Thinking Faster: Compressing LLM Chain-of-Thought via Step Entropy

Large Language Models (LLMs) using Chain-of-Thought (CoT) prompting excel at complex reasoning but generate verbose thought processes with considerable redundancy, leading to increased inference costs and reduced efficiency. We introduce a novel CoT compression framework based on step entropy, a metric that quantifies \emph{the informational contribution of individual reasoning steps} to identify redundancy. Through theoretical analysis and extensive empirical validation on mathematical reasoning benchmarks, we demonstrate that steps with low entropy are indeed highly redundant. Our experiments reveal that an astonishing 80\% of low-entropy intermediate steps can be pruned with minor degradation in the final answer accuracy across DeepSeek-R1-7B, 14B and Qwen3-8B. This finding sharply contrasts with random or high-entropy pruning, which severely impairs reasoning performance. Building on this, we propose a novel two-stage training strategy combining Supervised Fine-Tuning (SFT) and Group Relative Policy Optimization (GRPO) reinforcement learning. This approach enables LLMs to autonomously learn to generate compressed COTs during inference by strategically incorporating [SKIP] tokens. Our method significantly improves LLM inference efficiency while preserving accuracy, paving the way for more scalable LLM deployments and a better understanding of their internal reasoning. The code and data are released in https://github.com/staymylove/COT_Compresstion_via_Step_entropy.

cs.AI↗

arXiv2025

Reinforcement Learning with Verifiable Rewards Implicitly Incentivizes Correct Reasoning in Base LLMs

Recent advancements in long chain-of-thought (CoT) reasoning, particularly through the Group Relative Policy Optimization algorithm used by DeepSeek-R1, have led to significant interest in the potential of Reinforcement Learning with Verifiable Rewards (RLVR) for Large Language Models (LLMs). While RLVR promises to improve reasoning by allowing models to learn from free exploration, there remains debate over whether it truly enhances reasoning abilities or simply boosts sampling efficiency. This paper systematically investigates the impact of RLVR on LLM reasoning. We revisit Pass@K experiments and demonstrate that RLVR can extend the reasoning boundary for both mathematical and coding tasks. This is supported by our introduction of a novel evaluation metric, CoT-Pass@K, which captures reasoning success by accounting for both the final answer and intermediate reasoning steps. Furthermore, we present a theoretical framework explaining RLVR's incentive mechanism, demonstrating how it can encourage correct reasoning even when rewards are based solely on answer correctness. Our analysis of RLVR's training dynamics reveals that it incentivizes correct reasoning early in the process, with substantial improvements in reasoning quality confirmed through extensive evaluations. These findings provide strong evidence of RLVR's potential to enhance LLM reasoning, offering valuable insights into its mechanisms and performance improvements.

cs.AI↗

arXiv2025

Reasoning Scaffolding: Distilling the Flow of Thought from LLMs

The prevailing approach to distilling reasoning from Large Language Models (LLMs)-behavioral cloning from textual rationales-is fundamentally limited. It teaches Small Language Models (SLMs) to mimic surface-level patterns rather than the underlying algorithmic structure of thought, resulting in a critical lack of logical robustness. We argue that instead of cloning text, distillation should transfer this algorithmic structure directly. We introduce Reasoning Scaffolding}, a framework that reframes reasoning as a structured generation process. Our method first abstracts the teacher's thought process into a sequence of discrete, interpretable semantic signals (e.g., Contrast, Addition) that act as a scaffold. The student model is then trained via a multi-task objective to both (1)predict the next semantic signal, anticipating the reasoning flow, and (2)generate the corresponding step, conditioned on that signal. This multi-task scheme acts as a powerful regularizer, compelling the student to internalize the computational patterns of coherent reasoning. On a suite of challenging reasoning benchmarks, our method significantly outperforms state-of-the-art distillation in both accuracy and logical consistency, providing a path towards creating smaller models that are genuine reasoners, not just fluent mimics.

cs.AI↗

arXiv2025

AbGen: Evaluating Large Language Models in Ablation Study Design and Evaluation for Scientific Research

We introduce AbGen, the first benchmark designed to evaluate the capabilities of LLMs in designing ablation studies for scientific research. AbGen consists of 1,500 expert-annotated examples derived from 807 NLP papers. In this benchmark, LLMs are tasked with generating detailed ablation study designs for a specified module or process based on the given research context. Our evaluation of leading LLMs, such as DeepSeek-R1-0528 and o4-mini, highlights a significant performance gap between these models and human experts in terms of the importance, faithfulness, and soundness of the ablation study designs. Moreover, we demonstrate that current automated evaluation methods are not reliable for our task, as they show a significant discrepancy when compared to human assessment. To better investigate this, we develop AbGen-Eval, a meta-evaluation benchmark designed to assess the reliability of commonly used automated evaluation systems in measuring LLM performance on our task. We investigate various LLM-as-Judge systems on AbGen-Eval, providing insights for future research on developing more effective and reliable LLM-based evaluation systems for complex scientific tasks.

cs.CL↗

arXiv2025

Can LLMs Identify Critical Limitations within Scientific Research? A Systematic Evaluation on AI Research Papers

Peer review is fundamental to scientific research, but the growing volume of publications has intensified the challenges of this expertise-intensive process. While LLMs show promise in various scientific tasks, their potential to assist with peer review, particularly in identifying paper limitations, remains understudied. We first present a comprehensive taxonomy of limitation types in scientific research, with a focus on AI. Guided by this taxonomy, for studying limitations, we present LimitGen, the first comprehensive benchmark for evaluating LLMs' capability to support early-stage feedback and complement human peer review. Our benchmark consists of two subsets: LimitGen-Syn, a synthetic dataset carefully created through controlled perturbations of high-quality papers, and LimitGen-Human, a collection of real human-written limitations. To improve the ability of LLM systems to identify limitations, we augment them with literature retrieval, which is essential for grounding identifying limitations in prior scientific findings. Our approach enhances the capabilities of LLM systems to generate limitations in research papers, enabling them to provide more concrete and constructive feedback.

cs.CL↗

arXiv2025

Mixture of Low Rank Adaptation with Partial Parameter Sharing for Time Series Forecasting

Multi-task forecasting has become the standard approach for time-series forecasting (TSF). However, we show that it suffers from an Expressiveness Bottleneck, where predictions at different time steps share the same representation, leading to unavoidable errors even with optimal representations. To address this issue, we propose a two-stage framework: first, pre-train a foundation model for one-step-ahead prediction; then, adapt it using step-specific LoRA modules.This design enables the foundation model to handle any number of forecast steps while avoiding the expressiveness bottleneck. We further introduce the Mixture-of-LoRA (MoLA) model, which employs adaptively weighted LoRA experts to achieve partial parameter sharing across steps. This approach enhances both efficiency and forecasting performance by exploiting interdependencies between forecast steps. Experiments show that MoLA significantly improves model expressiveness and outperforms state-of-the-art time-series forecasting methods. Code is available at https://anonymous.4open.science/r/MoLA-BC92.

cs.LG↗

arXiv2025

Solve-Detect-Verify: Inference-Time Scaling with Flexible Generative Verifier

Large Language Model (LLM) reasoning for complex tasks inherently involves a trade-off between solution accuracy and computational efficiency. The subsequent step of verification, while intended to improve performance, further complicates this landscape by introducing its own challenging trade-off: sophisticated Generative Reward Models (GenRMs) can be computationally prohibitive if naively integrated with LLMs at test-time, while simpler, faster methods may lack reliability. To overcome these challenges, we introduce FlexiVe, a novel generative verifier that flexibly balances computational resources between rapid, reliable fast thinking and meticulous slow thinking using a Flexible Allocation of Verification Budget strategy. We further propose the Solve-Detect-Verify pipeline, an efficient inference-time scaling framework that intelligently integrates FlexiVe, proactively identifying solution completion points to trigger targeted verification and provide focused solver feedback. Experiments show FlexiVe achieves superior accuracy in pinpointing errors within reasoning traces on ProcessBench. Furthermore, on challenging mathematical reasoning benchmarks (AIME 2024, AIME 2025, and CNMO), our full approach outperforms baselines like self-consistency in reasoning accuracy and inference efficiency. Our system offers a scalable and effective solution to enhance LLM reasoning at test time.

cs.AI↗

arXiv2025

Intervention-Aware Forecasting: Breaking Historical Limits from a System Perspective

Traditional time series forecasting methods predominantly rely on historical data patterns, neglecting external interventions that significantly shape future dynamics. Through control-theoretic analysis, we show that the implicit "self-stimulation" assumption limits the accuracy of these forecasts. To overcome this limitation, we propose an Intervention-Aware Time Series Forecasting (IATSF) framework explicitly designed to incorporate external interventions. We particularly emphasize textual interventions due to their unique capability to represent qualitative or uncertain influences inadequately captured by conventional exogenous variables. We propose a leak-free benchmark composed of temporally synchronized textual intervention data across synthetic and real-world scenarios. To rigorously evaluate IATSF, we develop FIATS, a lightweight forecasting model that integrates textual interventions through Channel-Aware Adaptive Sensitivity Modeling (CASM) and Channel-Aware Parameter Sharing (CAPS) mechanisms, enabling the model to adjust its sensitivity to interventions and historical data in a channel-specific manner. Extensive empirical evaluations confirm that FIATS surpasses state-of-the-art methods, highlighting that forecasting improvements stem explicitly from modeling external interventions rather than increased model complexity alone.

cs.LG↗

arXiv2025

Dyve: Thinking Fast and Slow for Dynamic Process Verification

We present Dyve, a dynamic process verifier that enhances reasoning error detection in large language models by integrating fast and slow thinking, inspired by Kahneman's Systems Theory. Dyve adaptively applies immediate token-level confirmation System 1 for straightforward steps and comprehensive analysis System 2 for complex ones. Leveraging a novel step-wise consensus-filtered process supervision technique, combining Monte Carlo estimation with LLM based evaluation, Dyve curates high-quality supervision signals from noisy data. Experimental results on ProcessBench and the MATH dataset confirm that Dyve significantly outperforms existing process-based verifiers and boosts performance in Best-of-N settings.

cs.AI↗

arXiv2025

MMVU: Measuring Expert-Level Multi-Discipline Video Understanding

We introduce MMVU, a comprehensive expert-level, multi-discipline benchmark for evaluating foundation models in video understanding. MMVU includes 3,000 expert-annotated questions spanning 27 subjects across four core disciplines: Science, Healthcare, Humanities & Social Sciences, and Engineering. Compared to prior benchmarks, MMVU features three key advancements. First, it challenges models to apply domain-specific knowledge and perform expert-level reasoning to analyze specialized-domain videos, moving beyond the basic visual perception typically assessed in current video benchmarks. Second, each example is annotated by human experts from scratch. We implement strict data quality controls to ensure the high quality of the dataset. Finally, each example is enriched with expert-annotated reasoning rationals and relevant domain knowledge, facilitating in-depth analysis. We conduct an extensive evaluation of 32 frontier multimodal foundation models on MMVU. The latest System-2-capable models, o1 and Gemini 2.0 Flash Thinking, achieve the highest performance among the tested models. However, they still fall short of matching human expertise. Through in-depth error analyses and case studies, we offer actionable insights for future advancements in expert-level, knowledge-intensive video understanding for specialized domains.

cs.CV↗

arXiv2024

Preliminary Study on Incremental Learning for Large Language Model-based Recommender Systems

Adapting Large Language Models for Recommendation (LLM4Rec) has shown promising results. However, the challenges of deploying LLM4Rec in real-world scenarios remain largely unexplored. In particular, recommender models need incremental adaptation to evolving user preferences, while the suitability of traditional incremental learning methods within LLM4Rec remains ambiguous due to the unique characteristics of Large Language Models (LLMs). In this study, we empirically evaluate two commonly employed incremental learning strategies (full retraining and fine-tuning) for LLM4Rec. Surprisingly, neither approach shows significant improvements in the performance of LLM4Rec. Instead of dismissing the role of incremental learning, we attribute the lack of anticipated performance enhancement to a mismatch between the LLM4Rec architecture and incremental learning: LLM4Rec employs a single adaptation module for learning recommendations, limiting its ability to simultaneously capture long-term and short-term user preferences in the incremental learning context. To test this speculation, we introduce a Long- and Short-term Adaptation-aware Tuning (LSAT) framework for incremental learning in LLM4Rec. Unlike the single adaptation module approach, LSAT utilizes two distinct adaptation modules to independently learn long-term and short-term user preferences. Empirical results verify that LSAT enhances performance, thereby validating our speculation. We release our code at: https://github.com/TianhaoShi2001/LSAT.

cs.IR↗

arXiv2024

DSPO: An End-to-End Framework for Direct Sorted Portfolio Construction

In quantitative investment, constructing characteristic-sorted portfolios is a crucial strategy for asset allocation. Traditional methods transform raw stock data of varying frequencies into predictive characteristic factors for asset sorting, often requiring extensive manual design and misalignment between prediction and optimization goals. To address these challenges, we introduce Direct Sorted Portfolio Optimization (DSPO), an innovative end-to-end framework that efficiently processes raw stock data to construct sorted portfolios directly. DSPO's neural network architecture seamlessly transitions stock data from input to output while effectively modeling the intra-dependency of time-steps and inter-dependency among all tradable stocks. Additionally, we incorporate a novel Monotonical Logistic Regression loss, which directly maximizes the likelihood of constructing optimal sorted portfolios. To the best of our knowledge, DSPO is the first method capable of handling market cross-sections with thousands of tradable stocks fully end-to-end from raw multi-frequency data. Empirical results demonstrate DSPO's effectiveness, yielding a RankIC of 10.12\% and an accumulated return of 121.94\% on the New York Stock Exchange in 2023-2024, and a RankIC of 9.11\% with a return of 108.74\% in other markets during 2021-2022.

q-fin.PM↗

arXiv2024

Multi-Patch Prediction: Adapting LLMs for Time Series Representation Learning

In this study, we present aLLM4TS, an innovative framework that adapts Large Language Models (LLMs) for time-series representation learning. Central to our approach is that we reconceive time-series forecasting as a self-supervised, multi-patch prediction task, which, compared to traditional contrastive learning or mask-and-reconstruction methods, captures temporal dynamics in patch representations more effectively. Our strategy encompasses two-stage training: (i). a causal continual pre-training phase on various time-series datasets, anchored on next patch prediction, effectively syncing LLM capabilities with the intricacies of time-series data; (ii). fine-tuning for multi-patch prediction in the targeted time-series context. A distinctive element of our framework is the patch-wise decoding layer, which departs from previous methods reliant on sequence-level decoding. Such a design directly transposes individual patches into temporal sequences, thereby significantly bolstering the model's proficiency in mastering temporal patch-based representations. aLLM4TS demonstrates superior performance in several downstream tasks, proving its effectiveness in deriving temporal representations with enhanced transferability and marking a pivotal advancement in the adaptation of LLMs for time-series analysis.

cs.LG↗

arXiv2024

FITS: Modeling Time Series with $10k$ Parameters

In this paper, we introduce FITS, a lightweight yet powerful model for time series analysis. Unlike existing models that directly process raw time-domain data, FITS operates on the principle that time series can be manipulated through interpolation in the complex frequency domain. By discarding high-frequency components with negligible impact on time series data, FITS achieves performance comparable to state-of-the-art models for time series forecasting and anomaly detection tasks, while having a remarkably compact size of only approximately $10k$ parameters. Such a lightweight model can be easily trained and deployed in edge devices, creating opportunities for various applications. The code is available in: \url{https://github.com/VEWOXIC/FITS}

cs.LG↗

arXiv2023

The Stellar Abundances and Galactic Evolution Survey (SAGES) -- -- I. General Description and the First Data Release (DR1)

The Stellar Abundances and Galactic Evolution Survey (SAGES) of the northern sky is a specifically-designed multi-band photometric survey aiming to provide reliable stellar parameters with accuracy comparable to those from low-resolution optical spectra. It was carried out with the 2.3-m Bok telescope of Steward Observatory and three other telescopes. The observations in the $u_s$ and $v_s$ passband produced over 36,092 frames of images in total, covering a sky area of $\sim9960$ degree$^2$. The median survey completeness of all observing fields for the two bands are of $u_{\rm s}=20.4$ mag and $v_s=20.3$ mag, respectively, while the limiting magnitudes with signal-to-noise ratio (S/N) of 100 are $u_s\sim17$ mag and $v_s\sim18$ mag, correspondingly. We combined our catalog with the data release 1 (DR1) of the first of Panoramic Survey Telescope And Rapid Response System (Pan-STARRS1, PS1) catalog, and obtained a total of 48,553,987 sources which have at least one photometric measurement in each of the SAGES $u_s$ and $v_s$ and PS1 $grizy$ passbands, which is the DR1 of SAGES and it will be released in our paper. We compare our $gri$ point-source photometry with those of PS1 and found an RMS scatter of $\sim2$% in difference of PS1 and SAGES for the same band. We estimated an internal photometric precision of SAGES to be on the order of $\sim1$%. Astrometric precision is better than $0^{\prime\prime}.2$ based on comparison with the DR1 of Gaia mission. In this paper, we also describe the final end-user database, and provide some science applications.

astro-ph.GA↗