Researcher — paper.dinq.me

2026 顶会论文

作者位次 #3/3ICML 2026Poster· Theory / Learning Theory

The Implicit Bias of Depth: From Neural Collapse to Softmax Codes

Connall Garrod · Jonathan Keating · Christos Thrampoulidis

社会与人本理论与学习动力学OpenReview

作者位次 #8/9ICML 2026Poster· Deep Learning / Large Language Models

Scaling Generative Verifiers For Natural Language Mathematical Proof Verification And Selection

Sadegh Mahdavi · Branislav Kisacanin · Shubham Toshniwal · Wei Du · Ivan Moshkov · George Armstrong · Renjie Liao · Christos Thrampoulidis · Igor Gitman

推理OpenReview

作者位次 #3/3ICML 2026Poster

Asymmetric Prompt Weighting for Reinforcement Learning with Verifiable Rewards

Reinhard Heckel · Mahdi Soltanolkotabi · Christos Thrampoulidis

强化学习与后训练OpenReview

作者位次 #5/6ICML 2026Poster· Reinforcement Learning / Deep RL

On Group Relative Policy Optimization Collapse in Agent Search: The Lazy Likelihood-Displacement

wenlong deng · Yushu Li · Boying Gong · YI REN · Christos Thrampoulidis · Xiaoxiao Li

强化学习与后训练智能体与工具检索与知识OpenReview

作者位次 #3/3ICML 2026Poster· Deep Learning / Large Language Models

Facts in Stats: Impacts of Pretraining Diversity on Language Model Generalization

Tina Behnia · Puneesh Deora · Christos Thrampoulidis

数据与预训练检索与知识理论与学习动力学OpenReview

作者位次 #6/8ACL 2026Poster· Natural Language Processing

For-Value: Efficient Forward-Only Data Valuation for finetuning LLM s and VLM s

Wenlong Deng · Qi Zeng · Jiaming Zhang · Minghui Chen · Zixin Ding · Christos Thrampoulidis · Boying Gong · Xiaoxiao Li

摘要

Data valuation is essential for enhancing the transparency and accountability of large language models (LLMs) and vision-language models (VLMs). However, existing methods typically rely on gradient computations, making them computationally prohibitive for billion-parameter models and precluding b...

多模态理论与学习动力学对齐与偏好OpenReview 项目页

Christos Thrampoulidis

获取完整研究档案

联系方式