Imagen de Portada

LogitTrace: Detecting Benchmark Contamination via Layerwise Logit Trajectories

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	He, Zirui, Zhao, Haiyan, Li, Yingcong, Payani, Ali, du, Mengnan
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

Universal Activation Verbalizer: A Unified Framework for Cross-Model Activation Explanation
por: Zhao, Haiyan, et al.
Publicado: (2026)

LogitDynamics: Reliable ViT Error Detection from Layerwise Logit Trajectories
por: Beigelman, Ido, et al.
Publicado: (2026)

Rep2Text: Decoding Full Text from a Single LLM Token Representation
por: Zhao, Haiyan, et al.
Publicado: (2025)

SAIF: A Sparse Autoencoder Framework for Interpreting and Steering Instruction Following of Language Models
por: He, Zirui, et al.
Publicado: (2025)

SAE-SSV: Supervised Steering in Sparse Representation Spaces for Reliable Control of Language Models
por: He, Zirui, et al.
Publicado: (2025)

VQ-Logits: Compressing the Output Bottleneck of Large Language Models via Vector Quantized Logits
por: Shao, Jintian, et al.
Publicado: (2025)

Logit Reweighting for Topic-Focused Summarization
por: Braun, Joschka, et al.
Publicado: (2025)

LogitsCoder: Towards Efficient Chain-of-Thought Path Search via Logits Preference Decoding for Code Generation
por: Chen, Jizheng, et al.
Publicado: (2026)

Beyond Single Concept Vector: Modeling Concept Subspace in LLMs with Gaussian Distribution
por: Zhao, Haiyan, et al.
Publicado: (2024)

DALD: Improving Logits-based Detector without Logits from Black-box LLMs
por: Zeng, Cong, et al.
Publicado: (2024)

LogitLens4LLMs: Extending Logit Lens Analysis to Modern Large Language Models
por: Wang, Zhenyu
Publicado: (2025)

Spectral Logit Sculpting: Adaptive Low-Rank Logit Transformation for Controlled Text Generation
por: Li, Jin, et al.
Publicado: (2025)

ROME: Memorization Insights from Text, Logits and Representation
por: Li, Bo, et al.
Publicado: (2024)

Stabilizing Policy Optimization via Logits Convexity
por: Chen, Hongzhan, et al.
Publicado: (2026)

Large Vision-Language Model Alignment and Misalignment: A Survey Through the Lens of Explainability
por: Shu, Dong, et al.
Publicado: (2025)

DeLTa: A Decoding Strategy based on Logit Trajectory Prediction Improves Factuality and Reasoning Ability
por: He, Yunzhen, et al.
Publicado: (2025)

OmniOPD: Logit-Free On-Policy Distillation via Speculative Verification
por: Zhou, Yuhang, et al.
Publicado: (2026)

Language Ranker: A Metric for Quantifying LLM Performance Across High and Low-Resource Languages
por: Li, Zihao, et al.
Publicado: (2024)

Beyond Logit Lens: Contextual Embeddings for Robust Hallucination Detection & Grounding in VLMs
por: Phukan, Anirudh, et al.
Publicado: (2024)

LoCa: Logit Calibration for Knowledge Distillation
por: Yang, Runming, et al.
Publicado: (2024)

Limits of n-gram Style Control for LLMs via Logit-Space Injection
por: Ahmed, Sami-ul
Publicado: (2026)

LogitSpec: Accelerating Retrieval-based Speculative Decoding via Next Next Token Speculation
por: Liu, Tianyu, et al.
Publicado: (2025)

Logit Arithmetic Elicits Long Reasoning Capabilities Without Training
por: Zhang, Yunxiang, et al.
Publicado: (2025)

Understanding GUI Agent Localization Biases through Logit Sharpness
por: Tao, Xingjian, et al.
Publicado: (2025)

DistillLens: Symmetric Knowledge Distillation Through Logit Lens
por: Dhakal, Manish, et al.
Publicado: (2026)

InfiGFusion: Graph-on-Logits Distillation via Efficient Gromov-Wasserstein for Model Fusion
por: Wang, Yuanyi, et al.
Publicado: (2025)

Efficient Low-Resource Language Adaptation via Multi-Source Dynamic Logit Fusion
por: Zhang, Chen, et al.
Publicado: (2026)

Lost in Overlap: Exploring Logit-based Watermark Collision in LLMs
por: Luo, Yiyang, et al.
Publicado: (2024)

Logit-Entropy Adaptive Stopping Heuristic for Efficient Chain-of-Thought Reasoning
por: Quamar, Mohammad Atif, et al.
Publicado: (2025)

Think in Parallel, Answer as One: Logit Averaging for Open-Ended Reasoning
por: Wang, Haonan, et al.
Publicado: (2025)

Chain-of-Thought Augmentation with Logit Contrast for Enhanced Reasoning in Language Models
por: Shim, Jay, et al.
Publicado: (2024)

Towards Cross-Tokenizer Distillation: the Universal Logit Distillation Loss for LLMs
por: Boizard, Nicolas, et al.
Publicado: (2024)

Logit Arithmetic Elicits Long Reasoning Capabilities Without Training
por: Zhang, Yunxiang, et al.
Publicado: (2025)

Steering Language Models Before They Speak: Logit-Level Interventions
por: An, Hyeseon, et al.
Publicado: (2026)

Strategic Deflection: Defending LLMs from Logit Manipulation
por: Rachidy, Yassine, et al.
Publicado: (2025)

Logits are All We Need to Adapt Closed Models
por: Hiranandani, Gaurush, et al.
Publicado: (2025)

SSG: Logit-Balanced Vocabulary Partitioning for LLM Watermarking
por: Gu, Chenxi, et al.
Publicado: (2026)

Exploiting LLMs for Automatic Hypothesis Assessment via a Logit-Based Calibrated Prior
por: Gong, Yue, et al.
Publicado: (2025)

Last Layer Logits to Logic: Empowering LLMs with Logic-Consistent Structured Knowledge Reasoning
por: Li, Songze, et al.
Publicado: (2025)

Transferring Extreme Subword Style Using Ngram Model-Based Logit Scaling
por: Messner, Craig, et al.
Publicado: (2025)