Institution profile

HiThink Research

Industry researchasia · cn
Official website
Research library34linked papers
Opportunities0open roles
Selected work

Representative Papers

PSC: Extending Context Window of Large Language Models via Phase Shift Calibration

May 18, 2025Conference on Empirical Methods in Natural Language Processing

To address the challenge of suboptimal preset frequency scaling factors in RoPE-based context window extension—where the search space grows exponentially—this paper proposes a lightweight Phase Shift Calibration (PSC) module. PSC introduces, for the first time, a learnable phase shift mechanism that dynamically calibrates preconfigured frequency scaling (e.g., in PI, YaRN, and LongRoPE) without altering the original RoPE architecture or requiring model retraining. Its core components include differentiable phase calibration, a lightweight linear projection, and context-length-adaptive initialization. Experiments demonstrate that PSC consistently reduces perplexity on long-context benchmarks (16K–64K tokens), exhibits robust cross-model (Llama, Qwen) and cross-task (QA, long-document reasoning) performance, and delivers increasingly substantial gains as context length grows. Overall, PSC significantly enhances the robustness and plug-and-play applicability of existing RoPE extension methods.

1 citationsRead paper

MMDynOpt-Agent: Dynamic Optimization for Multimodal Large Language Model Reasoning via Reinforcement Learning

Aug 14, 2026

This study addresses the inefficient alignment between visual cues and question semantics in multimodal large language models by formulating multimodal reasoning as a Markov Decision Process. We propose a reinforcement learning-based dynamic prompt optimization framework that employs a lightweight agent for end-to-end training. A budget-aware composite reward mechanism is designed to balance accuracy and efficiency, adaptively guiding the target model through multi-turn reasoning. Experimental results across 15 datasets demonstrate that our method significantly outperforms existing baselines, exhibiting strong generalization and cross-model transferability. Ultimately, this approach effectively achieves both efficient and precise multimodal reasoning, overcoming critical limitations in current semantic translation processes within multimodal systems.

0 citationsRead paper
Recent publications

Latest Papers

MMDynOpt-Agent: Dynamic Optimization for Multimodal Large Language Model Reasoning via Reinforcement Learning

Aug 14, 2026

This study addresses the inefficient alignment between visual cues and question semantics in multimodal large language models by formulating multimodal reasoning as a Markov Decision Process. We propose a reinforcement learning-based dynamic prompt optimization framework that employs a lightweight agent for end-to-end training. A budget-aware composite reward mechanism is designed to balance accuracy and efficiency, adaptively guiding the target model through multi-turn reasoning. Experimental results across 15 datasets demonstrate that our method significantly outperforms existing baselines, exhibiting strong generalization and cross-model transferability. Ultimately, this approach effectively achieves both efficient and precise multimodal reasoning, overcoming critical limitations in current semantic translation processes within multimodal systems.

0 citationsRead paper

TIEM: Temporal Integration of Hypergraph Evidence and Skill Memory for Event-Driven Financial Forecasting

Aug 13, 2026

This work addresses the “evidence gap” in event-driven financial forecasting caused by training data contamination and temporal leakage. To tackle this issue, the authors propose TIEM, a timestamp-gated framework that constructs an Event-Evidence Hypergraph (EEH) for multi-level temporal retrieval, incorporates a Case-based Skill Memory (CSM) module with source-aware labels to store temporal reasoning skills, and employs a Heterogeneous Evidence-Experience Fusion Reasoning (HEFR) mechanism for prediction. To rigorously evaluate a model’s genuine temporal sensitivity, they introduce the FinPURE benchmark along with the Name-Date Probe method. Experimental results demonstrate that TIEM significantly outperforms existing approaches across five financial forecasting benchmarks, confirming its effectiveness in realistic temporal settings.

0 citationsRead paper