Imagen de Portada

Understanding and Mitigating Numerical Sources of Nondeterminism in LLM Inference

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Yuan, Jiayi, Li, Hao, Ding, Xinheng, Xie, Wenya, Li, Yu-Jhe, Zhao, Wentian, Wan, Kun, Shi, Jing, Hu, Xia, Liu, Zirui
Formato:	Preprint
Publicado:	2025
Materias:	Computation and Language
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

DUMP: Automated Distribution-Level Curriculum Learning for RL-based LLM Post-training
por: Wang, Zhenting, et al.
Publicado: (2025)

Deterministic Inference across Tensor Parallel Sizes That Eliminates Training-Inference Mismatch
por: Zhang, Ziyang, et al.
Publicado: (2025)

Efficient Self-Improvement in Multimodal Large Language Models: A Model-Level Judge-Free Approach
por: Deng, Shijian, et al.
Publicado: (2024)

EPO: Entropy-regularized Policy Optimization for LLM Agents Reinforcement Learning
por: Xu, Wujiang, et al.
Publicado: (2025)

DiFR: Inference Verification Despite Nondeterminism
por: Karvonen, Adam, et al.
Publicado: (2025)

Treat Visual Tokens as Text? But Your MLLM Only Needs Fewer Efforts to See
por: Zhang, Zeliang, et al.
Publicado: (2024)

Living Arrangements and Women's Household Decision‐Making Power in China
por: Xinheng Li
Publicado: (2026)

noDice: Inference for Discrete Probabilistic Programs with Nondeterminism and Conditioning
por: Gürtler, Tobias, et al.
Publicado: (2026)

Resolving Nondeterminism by Chance
por: Paul, Soumyajit, et al.
Publicado: (2025)

A Hierarchy of Nondeterminism
por: Radi, Bader Abu, et al.
Publicado: (2022)

Resolving Nondeterminism with Randomness
por: Henzinger, Thomas A., et al.
Publicado: (2025)

Silencer: From Discovery to Mitigation of Self-Bias in LLM-as-Benchmark-Generator
por: Yuan, Peiwen, et al.
Publicado: (2025)

Probability and Angelic Nondeterminism with Multiset Semantics
por: Ong, Shawn, et al.
Publicado: (2024)

A Novel LLM-based Two-stage Summarization Approach for Long Dialogues
por: Yin, Yuan-Jhe, et al.
Publicado: (2024)

Link between cascade transitions and correlated Chern insulators in magic-angle twisted bilayer graphene
por: Hu, Qianying, et al.
Publicado: (2024)

“A Contribution Shared Is a Contribution Halved”: The Influence of Tie Strength‐Based Disclosure Strategy on Potential Donors' Donation Amounts
por: Jing Zhao, et al.
Publicado: (2025)

When Does Multi-Agent RL Improve LLM Workflows? Workflow, Scale, and Policy-Sharing Tradeoffs
por: Zeng, Yifan, et al.
Publicado: (2026)

STARE at the Structure: Steering ICL Exemplar Selection with Structural Alignment
por: Li, Jiaqian, et al.
Publicado: (2025)

Prompt-Guided Mask Proposal for Two-Stage Open-Vocabulary Segmentation
por: Li, Yu-Jhe, et al.
Publicado: (2024)

Language Equivalence is Undecidable in VASS with Restricted Nondeterminism
por: Czerwiński, Wojciech, et al.
Publicado: (2025)

A Demonic Outcome Logic for Randomized Nondeterminism
por: Zilberstein, Noam, et al.
Publicado: (2024)

Optimistic Verifiable Training by Controlling Hardware Nondeterminism
por: Srivastava, Megha, et al.
Publicado: (2024)

Beyond Numeric Rewards: In-Context Dueling Bandits with LLM Agents
por: Xia, Fanzeng, et al.
Publicado: (2024)

Capacity-Aware Inference: Mitigating the Straggler Effect in Mixture of Experts
por: He, Shwai, et al.
Publicado: (2025)

Vision-Zero: Scalable VLM Self-Improvement via Strategic Gamified Self-Play
por: Wang, Qinsi, et al.
Publicado: (2025)

ZID-Net: Zero-Inference Diffusion Prior Decoupling Network for Single Image Dehazing
por: Li, Xinheng, et al.
Publicado: (2026)

From High to Low: Simulating Nondeterminism and State with State
por: Tang, Wenhao, et al.
Publicado: (2023)

Nondeterminism makes unary 1-limited automata concise
por: Guillon, Bruno, et al.
Publicado: (2025)

AttnRegDeepLab: A Two-Stage Decoupled Framework for Interpretable Embryo Fragmentation Grading
por: Lee, Ming-Jhe
Publicado: (2025)

Quantitative lower bound for solutions to the Boltzmann equation in non-convex domains
por: Su, Jhe-Kuan
Publicado: (2025)

Characterization-Guided GPU Fault Resilience in NVIDIA MPS
por: Liu, Rixin, et al.
Publicado: (2026)

Understanding and Mitigating Errors of LLM-Generated RTL Code
por: Zhang, Jiazheng, et al.
Publicado: (2025)

NODE-AdvGAN: Improving the transferability and perceptual similarity of adversarial examples by dynamic-system-driven adversarial generative model
por: Xie, Xinheng, et al.
Publicado: (2024)

Generating Mitigations for Downstream Projects to Neutralize Upstream Library Vulnerability
por: Chen, Zirui, et al.
Publicado: (2025)

How Memory Management Impacts LLM Agents: An Empirical Study of Experience-Following Behavior
por: Xiong, Zidi, et al.
Publicado: (2025)

Multi-Functional RIS-Enabled in SAGIN for IoT: A Hybrid Deep Reinforcement Learning Approach with Compressed Twin-Models
por: Shen, Li-Hsiang, et al.
Publicado: (2025)

Vocabulary Hijacking in LVLMs: Unveiling Critical Attention Heads by Excluding Inert Tokens to Mitigate Hallucination
por: Chen, Yangneng, et al.
Publicado: (2026)

RL of Thoughts: Navigating LLM Reasoning with Inference-time Reinforcement Learning
por: Hao, Qianyue, et al.
Publicado: (2025)

When Continual Learning Moves to Memory: A Study of Experience Reuse in LLM Agents
por: Hu, Qisheng, et al.
Publicado: (2026)

A Bit of Nondeterminism Makes Pushdown Automata Expressive and Succinct
por: Guha, Shibashis, et al.
Publicado: (2021)