--- title: "Jie Zhou — papers, page 2 — Elamur Research" description: "Jie Zhou, Tencent: 41 papers on Elamur Research (2015–2026), h-index 29 — affiliations, topics, co-authors and who cites them in our library." canonical: "https://research.elamur.ai/free/en/authors/jie-zhou/page/2/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/authors/jie-zhou/page/2/) [home](https://md.research.elamur.ai/free/en/) / [authors](https://md.research.elamur.ai/free/en/authors/) / [Jie Zhou](https://md.research.elamur.ai/free/en/authors/jie-zhou/) / page 2 # Jie Zhou · page 2 41 papers in our library · 2015–2026 · page 2 of 2 [The profile, affiliations and co-authors →](https://md.research.elamur.ai/free/en/authors/jie-zhou/) ## Papers The most cited first (OpenAlex and our library). [← prev](https://md.research.elamur.ai/free/en/authors/jie-zhou/) [1](https://md.research.elamur.ai/free/en/authors/jie-zhou/) **2** 26. **[Think Natively: Unlocking Multilingual Reasoning with Consistency-Enhanced Reinforcement Learning](https://md.research.elamur.ai/free/en/papers/zhang-2026-think-natively-unlocking-multilingual-reasoning/)** *(2026 · 11 citations)* Reasoning-oriented LLMs that think before answering often handle non-English prompts poorly. They tend to drift into English or a mix of languages during reasoning, and their accuracy drops compared with English. The authors introduce M-Thinker, a family of models (1.5B, 4B, 7B) trained with GRPO using two new reward… 27. **[A Survey of Inductive Reasoning for Large Language Models](https://md.research.elamur.ai/free/en/papers/chen-2026-survey-inductive-reasoning-large-language/)** *(2026 · 10 citations)* Inductive reasoning moves from particular observations to general rules, and it often admits several valid answers. 28. **[Investigating Forgetting in Pre-Trained Representations Through Continual Learning](https://md.research.elamur.ai/free/en/papers/luo-2023-investigating-forgetting-pre-trained-representations/)** *(2023 · 10 citations)* When a pre-trained language model is fine-tuned on a sequence of tasks, its contextualized representations shift. 29. **[Cross-layer Attention Sharing for Pre-trained Large Language Models](https://md.research.elamur.ai/free/en/papers/mu-2026-cross-layer-attention-sharing-pre/)** *(2026 · 2 citations)* Earlier work on making LLM attention more efficient mostly targets redundancy inside a layer, for example by compressing the KV cache or grouping heads. 30. **[LLM-KT: Enhancing Large Language Models with Knowledge Tracing via Multi-Level Plug-and-Play Alignment](https://md.research.elamur.ai/free/en/papers/wang-2026-llm-kt-enhancing-large-language/)** *(2026 · 1 citation)* Knowledge Tracing (KT) estimates whether a learner will solve the next exercise from their earlier answers. 31. **[PsychEval: A Multi-Session and Multi-Therapy Benchmark for High-Realism AI Psychological Counselor](https://md.research.elamur.ai/free/en/papers/pan-2026-psycheval-multi-session-multi-therapy/)** *(2026 · 1 citation)* Most benchmarks for AI mental-health assistants test isolated, single-turn exchanges. Real counseling, by contrast, unfolds over many sessions, depends on remembering the client's history, and often mixes therapeutic schools. PsychEval addresses this gap with a Chinese-language benchmark built by using LLMs (e.g.… 32. **[TamEdit: Trajectory-Aware Meta-Learning for Specificity-Preserving Continual Knowledge Editing](https://md.research.elamur.ai/free/en/papers/tian-2026-tamedit-trajectory-aware-meta-learning/)** *(2026 · 1 citation)* The paper tackles a weakness of continual knowledge editing in large language models. When many factual updates are applied one after another, existing editors compute each change on its own, and unrelated knowledge gradually degrades. The authors call this specificity drift. They propose TamEdit, which treats a… 33. **[UltraEval-Audio: A Unified Framework for Comprehensive Evaluation of Audio Foundation Models](https://md.research.elamur.ai/free/en/papers/shi-2026-ultraeval-audio-unified-framework-comprehensive/)** *(2026 · 1 citation)* Evaluating audio foundation models (AFMs) is difficult because benchmarks and evaluation code are scattered across sources. 34. **[ArrowGEV: Grounding Events in Video via Learning the Arrow of Time](https://md.research.elamur.ai/free/en/papers/yu-2026-arrowgev-grounding-events-video-learning/)** *(2026)* The paper addresses video event grounding (GEV): finding the time span in an untrimmed video that matches a text description. 35. **[FeedTrans: Enhance Machine Translation via Feedback](https://md.research.elamur.ai/free/en/papers/sun-2026-feedtrans-enhance-machine-translation-feedback/)** *(2026)* Reinforcement learning (RL) for machine translation usually rewards a model only for what it can already produce. 36. **[FlowRAG: Synergizing Explicit Reasoning via Frequency-Aware Multi-Granularity Graph Flow](https://md.research.elamur.ai/free/en/papers/zhan-2026-flowrag-synergizing-explicit-reasoning-frequency/)** *(2026)* Graph-based RAG systems usually start retrieval from entities and spread relevance implicitly through the graph. 37. **[Investigating Cross-Modal Skill Injection: Scenarios, Methods, and Hyperparameters](https://md.research.elamur.ai/free/en/papers/xu-2026-investigating-cross-modal-skill-injection/)** *(2026)* Vision-language models (VLMs) handle general multimodal tasks well, but they are slow to acquire new specialized skills. 38. **[Joint Optimization of Training Data and Policy in RLHF](https://md.research.elamur.ai/free/en/papers/yu-2026-joint-optimization-training-data-policy/)** *(2026)* Standard RLHF and RLVR pipelines sample many responses per training prompt but never change the prompt itself. 39. **[More Than Sum of Its Parts: Deciphering Intent Shifts in Multimodal Hate Speech Detection](https://md.research.elamur.ai/free/en/papers/sun-2026-more-than-sum-its-parts/)** *(2026)* The paper addresses implicit multimodal hate speech, where a text and an image look harmless on their own but turn hateful together. 40. **[ReFreeKV: Towards Threshold-Free KV Cache Compression](https://md.research.elamur.ai/free/en/papers/ni-2026-refreekv-threshold-free-kv-cache/)** *(2026)* KV cache pruning methods such as H2O, SnapKV and CAKE can cut LLM inference memory with little loss, but only when the cache budget is tuned to each dataset or domain. 41. **[Situated Embedding Models for Context-Aware Dense Retrieval](https://md.research.elamur.ai/free/en/papers/wu-2026-situated-embedding-models-context-aware/)** *(2026)* When RAG systems split long documents into chunks, a chunk often cannot be interpreted correctly without the surrounding text. [← prev](https://md.research.elamur.ai/free/en/authors/jie-zhou/) [1](https://md.research.elamur.ai/free/en/authors/jie-zhou/) **2**