Author name cluster

Anqi Li

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

8 papers

2 author rows

AAAI Conference 2026 Conference Paper

ARBench: Algorithmic Reasoner or API Alchemist? Evaluating LLMs Beyond API Calls

Ren-Biao Liu
Chao-Zeng Ma
Anqi Li
Hui Sun
Xin-Ye Li
Ming Li

Large Language Models (LLMs) have demonstrated impressive capabilities in code generation. Like human programmers, LLMs tend to call high-level APIs and libraries to program efficiently. However, this shortcut may hinder LLMs from learning the essential algorithm reasoning, leading instead to rote memorization of API usage. As a result, LLMs often struggle to generalize to new or domain-specific algorithms that lack ready-made library support. In this work, we propose ARBench, a novel benchmark for evaluating LLMs’ ability to generate machine learning algorithms from scratch, beyond merely invoking high-level APIs. It emphasizes algorithmic reasoning and implementation, distinguishing genuine understanding from superficial API usage. It covers fundamental and advanced machine learning tasks, rigorously assessing current LLMs’ capacity to implement these algorithms from scratch. Our evaluation reveals the strengths and weaknesses of state-of-the-art LLMs in algorithmic reasoning and generalization, offering valuable insights to guide future research and development.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

FlowCut: Rethinking Redundancy via Information Flow for Efficient Vision-Language Models

Jintao Tong
Wenwei Jin
Pengda Qin
Anqi Li
Yixiong Zou
Yuhong Li
Yuhua Li
Ruixuan Li

Large vision-language models (LVLMs) excel at multimodal understanding but suffer from high computational costs due to redundant vision tokens. Existing pruning methods typically rely on single-layer attention scores to rank and prune redundant visual tokens to solve this inefficiency. However, as the interaction between tokens and layers is complicated, this raises a basic question: Is such a simple single-layer criterion sufficient to identify redundancy? To answer this question, we rethink the emergence of redundant visual tokens from a fundamental perspective: information flow, which models the interaction between tokens and layers by capturing how information moves between tokens across layers. We find (1) the CLS token acts as an information relay, which can simplify the complicated flow analysis; (2) the redundancy emerges progressively and dynamically via layer-wise attention concentration; and (3) relying solely on attention scores from single layers can lead to contradictory redundancy identification. Based on this, we propose FlowCut, an information-flow-aware pruning framework, mitigating the insufficiency of the current criterion for identifying redundant tokens and better aligning with the model's inherent behaviors. Extensive experiments show FlowCut achieves superior results, outperforming SoTA by 1. 6% on LLaVA-1. 5-7B with 88. 9% token reduction, and by 4. 3% on LLaVA-NeXT-7B with 94. 4% reduction, delivering 3. 2$\times$ speed-up in the prefilling stage. Our code is available at https: //github. com/TungChintao/FlowCut.

PDF Details

ICLR Conference 2025 Conference Paper

Once-for-All: Controllable Generative Image Compression with Dynamic Granularity Adaptation

Anqi Li
Feng Li 0037
Yuxi Liu 0020
Runmin Cong
Yao Zhao 0001
Huihui Bai 0001

Although recent generative image compression methods have demonstrated impressive potential in optimizing the rate-distortion-perception trade-off, they still face the critical challenge of flexible rate adaptation to diverse compression necessities and scenarios. To overcome this challenge, this paper proposes a $\textbf{Control}$lable $\textbf{G}$enerative $\textbf{I}$mage $\textbf{C}$ompression framework, $\textbf{Control-GIC}$, the first capable of fine-grained bitrate adaptation across a broad spectrum while ensuring high-fidelity and generality compression. We base Control-GIC on a VQGAN framework representing an image as a sequence of variable-length codes ($\textit{i.e.}$ VQ-indices), which can be losslessly compressed and exhibits a direct positive correlation with bitrates. Drawing inspiration from the classical coding principle, we correlate the information density of local image patches with their granular representations. Hence, we can flexibly determine a proper allocation of granularity for the patches to achieve dynamic adjustment for VQ-indices, resulting in desirable compression rates. We further develop a probabilistic conditional decoder capable of retrieving historic encoded multi-granularity representations according to transmitted codes, and then reconstruct hierarchical granular features in the formalization of conditional probability, enabling more informative aggregation to improve reconstruction realism. Our experiments show that Control-GIC allows highly flexible and controllable bitrate adaptation where the results demonstrate its superior performance over recent state-of-the-art methods.

Details

NeurIPS Conference 2025 Conference Paper

Purity Law for Neural Routing Problem Solvers with Enhanced Generalizability

Wenzhao Liu
Haoran Li
Congying Han
Zicheng Zhang
Anqi Li
Tiande Guo

Achieving generalization in neural approaches across different scales and distributions remains a significant challenge for routing problems. A key obstacle is that neural networks often fail to learn robust principles for identifying universal patterns and deriving optimal solutions from diverse instances. In this paper, we first uncover Purity Law, a fundamental structural principle for optimal solutions of routing problems, defining that edge prevalence grows exponentially with the sparsity of surrounding vertices. Statistically and theoretically validated across diverse instances, Purity Law reveals a consistent bias toward local sparsity in global optima. Building on this insight, we propose Purity Policy Optimization (PUPO), a novel training paradigm that explicitly aligns characteristics of neural solutions with Purity Law during the solution construction process to enhance generalization. Extensive experiments demonstrate that PUPO can be seamlessly integrated with popular neural solvers, significantly enhancing their generalization performance without incurring additional computational overhead during inference.

PDF Details

ICML Conference 2025 Conference Paper

Revisiting Chain-of-Thought in Code Generation: Do Language Models Need to Learn Reasoning before Coding?

Renbiao Liu
Anqi Li
Chaoding Yang
Hui Sun 0003
Ming Li 0005

Large Language Models (LLMs) have demonstrated exceptional performance in code generation, becoming increasingly vital for software engineering and development. Recently, Chain-of-Thought (CoT) has proven effective for complex tasks by prompting LLMs to reason step-by-step and provide a final answer. However, research on how LLMs learn to reason with CoT data for code generation remains limited. In this work, we revisit classic CoT training, which typically learns reasoning steps before the final answer. We synthesize a dataset to separate the CoT process from code solutions and then conduct extensive experiments to study how CoT works in code generation empirically. We observe counterintuitive phenomena, suggesting that the traditional training paradigm may not yield benefits for code generation. Instead, training LLMs to generate code first and then output the CoT to explain reasoning steps for code generation is more effective. Specifically, our results indicate that a 9. 86% relative performance improvement can be achieved simply by changing the order between CoT and code. Our findings provide valuable insights into leveraging CoT to enhance the reasoning capabilities of CodeLLMs and improve code generation.

Details

SODA Conference 2023 Conference Paper

Cubic Goldreich-Levin

Dain Kim
Anqi Li
Jonathan Tidor

In this paper, we give a cubic Goldreich-Levin algorithm which makes polynomially-many queries to a function f: 𝔽 n p → ℂ and produces a decomposition of f as a sum of cubic phases and a small error term. This is a natural higher-order generalization of the classical Goldreich-Levin algorithm. The classical (linear) Goldreich-Levin algorithm has wide-ranging applications in learning theory, coding theory and the construction of pseudorandom generators in cryptography, as well as being closely related to Fourier analysis. Higher-order Goldreich-Levin algorithms on the other hand involve central problems in higher-order Fourier analysis, namely the inverse theory of the Gowers U k norms, which are well-studied in additive combinatorics. The only known result in this direction prior to this work is the quadratic Goldreich-Levin theorem, proved by Tulsiani and Wolf in 2011. The main step of their result involves an algorithmic version of the U 3 inverse theorem. More complications appear in the inverse theory of the U 4 and higher norms. Our cubic Goldreich-Levin algorithm is based on algorithmizing recent work by Gowers and Milicevic who proved new quantitative bounds for the U 4 inverse theorem. Our cubic Goldreich-Levin algorithm is constructed from two main tools: an algorithmic U 4 inverse theorem and an arithmetic decomposition result in the style of the Frieze-Kannan graph regularity lemma. As one application of our main theorem we solve the problem of self-correction for cubic Reed-Muller codes beyond the list decoding radius. Additionally we give a purely combinatorial result: an improvement of the quantitative bounds on the U 4 inverse theorem.

Details

NeurIPS Conference 2023 Conference Paper

Survival Instinct in Offline Reinforcement Learning

Anqi Li
Dipendra Misra
Andrey Kolobov
Ching-An Cheng

We present a novel observation about the behavior of offline reinforcement learning (RL) algorithms: on many benchmark datasets, offline RL can produce well-performing and safe policies even when trained with "wrong" reward labels, such as those that are zero everywhere or are negatives of the true rewards. This phenomenon cannot be easily explained by offline RL's return maximization objective. Moreover, it gives offline RL a degree of robustness that is uncharacteristic of its online RL counterparts, which are known to be sensitive to reward design. We demonstrate that this surprising robustness property is attributable to an interplay between the notion of pessimism in offline RL algorithms and certain implicit biases in common data collection practices. As we prove in this work, pessimism endows the agent with a survival instinct, i. e. , an incentive to stay within the data support in the long term, while the limited and biased data coverage further constrains the set of survival policies. Formally, given a reward class -- which may not even contain the true reward -- we identify conditions on the training data distribution that enable offline RL to learn a near-optimal and safe policy from any reward within the class. We argue that the survival instinct should be taken into account when interpreting results from existing offline RL benchmarks and when creating future ones. Our empirical and theoretical results suggest a new paradigm for offline RL, whereby an agent is "nudged" to learn a desirable behavior with imperfect reward but purposely biased data coverage. Please visit our website https: //survival-instinct. github. io for accompanied code and videos.

PDF Details

YNIMG Journal 2021 Journal Article

The coupling of BOLD signal variability and degree centrality underlies cognitive functions and psychiatric diseases

Jintao Sheng
Liang Zhang
Junjiao Feng
Jing Liu
Anqi Li
Wei Chen
Yuedi Shen
Jinhui Wang

Brain signal variability has been consistently linked to functional integration; however, whether this coupling is associated with cognitive functions and/or psychiatric diseases has not been clarified. Using multiple multimodality datasets, including resting-state functional magnetic resonance imaging (rsfMRI) data from the Human Connectome Project (HCP: N = 927) and a Beijing sample (N = 416) and cerebral blood flow (CBF) and rsfMRI data from a Hangzhou sample (N = 29), we found that, compared with the existing variability measure (i. e. , SDBOLD), the mean-scaled (standardized) fractional standard deviation of the BOLD signal (mfSDBOLD) maintained very high test-retest reliability, showed greater cross-site reliability and was less affected by head motion. We also found strong reproducible couplings between the mfSDBOLD and functional integration measured by the degree centrality (DC), both cross-voxel and cross-subject, which were robust to scanning and preprocessing parameters. Moreover, both mfSDBOLD and DC were correlated with CBF, suggesting a common physiological basis for both measures. Critically, the degree of coupling between mfSDBOLD and long-range DC was positively correlated with individuals’ cognitive total composite scores. Brain regions with greater mismatches between mfSDBOLD and long-range DC were more vulnerable to brain diseases. Our results suggest that BOLD signal variability could serve as a meaningful index of local function that underlies functional integration in the human brain and that a strong coupling between BOLD signal variability and functional integration may serve as a hallmark of balanced brain networks that are associated with optimal brain functions.

Details DOI