Large Language Diffusion Models (2025 · 178 citations)
Modern LLMs almost all rely on autoregressive next-token prediction. Many assume that scaling, in-context learning and instruction-following depend on this design. The authors test that assumption with LLaDA, a masked diffusion language model trained from scratch.
LLaDA 1.5: Variance-Reduced Preference Optimization for Large Language Diffusion Models (2026 · 49 citations)
Masked diffusion models (MDMs) for language cannot compute exact log-likelihoods, so preference alignment methods such as DPO cannot be applied directly.
Language Models Can Improve Event Prediction by Few-Shot Abductive Reasoning (2023 · 44 citations)
The paper asks whether large language models can help forecast real-world events in time-stamped sequences, such as political interactions between G20 entities, where each event carries text like a news headline.
Rethinking Sample Polarity in Reinforcement Learning with Verifiable Rewards (2026 · 4 citations)
This paper asks what positive (correct) and negative (incorrect) rollouts each contribute when large reasoning models are trained with RLVR.
ReportLogic: Evaluating Logical Quality in Deep Research Reports (2026 · 1 citation)
LLM-based Deep Research tools produce long analytical reports. Current evaluation mostly checks atomic facts, citations and fluency, so it misses whether the argument holds together and whether conclusions are actually backed. ReportLogic treats logical quality as auditability: can a reader follow, understand and…
SepSeq: A Training-Free Framework for Long Numerical Sequence Processing in LLMs (2026 · 1 citation)
Modern LLMs advertise very long context windows, yet they handle long sequences of numbers poorly. Tasks such as counting, filtering, or spotting trends need exact access to every value. The authors trace this failure to attention dispersion: as inputs grow, Softmax spreads attention over many similar numeric tokens…
Learning to Translate by Translating: Stabilizing the Dual Loop via Semantic-Aware Self-Evolution (2026)
Supervised LLM translation systems such as Tower and X-ALMA rely on large, expensive human-made parallel corpora.
RV-Syn: Rational and Verifiable Mathematical Reasoning Data Synthesis based on Structured Function Library (2026)
RV-Syn tackles the shortage of reliable, complex mathematical training data for LLMs. Prior synthesis methods either rework existing annotated sets such as GSM8K and MATH, which limits diversity, or have a model write questions directly from documents or knowledge points. The authors argue that direct generation…
Training-Free Adaptive Speculative Decoding via Linguistic Priors (2026)
Speculative decoding speeds up LLM inference: a small drafter proposes tokens and a large target model checks them in parallel.
UniGeM: Unifying Data Selection and Mixing via Geometric Exploration and Mining (2026)
Data quality now limits LLM pre-training more than data volume. Domain mixing and sample selection are usually handled as separate steps, which can damage the logical and hierarchical structure of code data. UniGeM combines both steps in one geometric pipeline. It needs no proxy model and no external reference…

Co-authors here

12 of 29 co-authors with a page, the most papers together first.

Venues

Where their works here were published, the most works first, with the editions of those that have a page.

Cited by

Papers of our library cite their works 226 times (of the 149,439 papers whose bibliographies we read); 37 citing papers with a page here →.

Updated 1 November 2026

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/authors/jun-zhou/

🧑‍🔬 Jun Zhou

🏛️ Affiliations

🧭 Fields, topics and concepts

📅 Works by year

2020s

📄 Papers

🤝 Co-authors here

Venues

🔗 Cited by

Jun Zhou

Affiliations

Fields, topics and concepts

Works by year

Papers

Co-authors here

Cited by