Image de couverture de livre

Limits of n-gram Style Control for LLMs via Logit-Space Injection

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteur principal:	Ahmed, Sami-ul
Format:	Preprint
Publié:	2026
Sujets:	Computation and Language
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Infini-gram mini: Exact n-gram Search at the Internet Scale with FM-Index
par: Xu, Hao, et autres
Publié: (2025)

The Role of $n$-gram Smoothing in the Age of Neural Networks
par: Malagutti, Luca, et autres
Publié: (2024)

Can Transformers Learn $n$-gram Language Models?
par: Svete, Anej, et autres
Publié: (2024)

Lngram: N-gram Conditional Memory in Latent Space
par: Zheng, Yunao, et autres
Publié: (2026)

Fingerprinting LLMs via Prompt Injection
par: Hu, Yuepeng, et autres
Publié: (2025)

Infini-gram: Scaling Unbounded n-gram Language Models to a Trillion Tokens
par: Liu, Jiacheng, et autres
Publié: (2024)

Transformers Can Represent $n$-gram Language Models
par: Svete, Anej, et autres
Publié: (2024)

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

Transferring Extreme Subword Style Using Ngram Model-Based Logit Scaling
par: Messner, Craig, et autres
Publié: (2025)

Beyond Prompting: Efficient and Robust Contextual Biasing for Speech LLMs via Logit-Space Integration (LOGIC)
par: Wang, Peidong
Publié: (2026)

LogitTrace: Detecting Benchmark Contamination via Layerwise Logit Trajectories
par: He, Zirui, et autres
Publié: (2025)

HITgram: A Platform for Experimenting with n-gram Language Models
par: Dasgupta, Shibaranjani, et autres
Publié: (2024)

DALD: Improving Logits-based Detector without Logits from Black-box LLMs
par: Zeng, Cong, et autres
Publié: (2024)

VQ-Logits: Compressing the Output Bottleneck of Large Language Models via Vector Quantized Logits
par: Shao, Jintian, et autres
Publié: (2025)

Towards Cross-Tokenizer Distillation: the Universal Logit Distillation Loss for LLMs
par: Boizard, Nicolas, et autres
Publié: (2024)

Lost in Overlap: Exploring Logit-based Watermark Collision in LLMs
par: Luo, Yiyang, et autres
Publié: (2024)

Identifying Influential N-grams in Confidence Calibration via Regression Analysis
par: Ozaki, Shintaro, et autres
Publié: (2026)

Modelling Intertextuality with N-gram Embeddings
par: Xing, Yi
Publié: (2025)

Exploiting LLMs for Automatic Hypothesis Assessment via a Logit-Based Calibrated Prior
par: Gong, Yue, et autres
Publié: (2025)

State space models can express n-gram languages
par: Nandakumar, Vinoth, et autres
Publié: (2023)

LogitsCoder: Towards Efficient Chain-of-Thought Path Search via Logits Preference Decoding for Code Generation
par: Chen, Jizheng, et autres
Publié: (2026)

Understanding Transformers via N-gram Statistics
par: Nguyen, Timothy
Publié: (2024)

Style-Specific Neurons for Steering LLMs in Text Style Transfer
par: Lai, Wen, et autres
Publié: (2024)

Spectral Logit Sculpting: Adaptive Low-Rank Logit Transformation for Controlled Text Generation
par: Li, Jin, et autres
Publié: (2025)

Last Layer Logits to Logic: Empowering LLMs with Logic-Consistent Structured Knowledge Reasoning
par: Li, Songze, et autres
Publié: (2025)

Strategic Deflection: Defending LLMs from Logit Manipulation
par: Rachidy, Yassine, et autres
Publié: (2025)

From n-gram to Attention: How Model Architectures Learn and Propagate Bias in Language Modeling
par: Kabir, Mohsinul, et autres
Publié: (2025)

Efficient Knowledge Injection in LLMs via Self-Distillation
par: Kujanpää, Kalle, et autres
Publié: (2024)

From Style to Facts: Mapping the Boundaries of Knowledge Injection with Finetuning
par: Zhao, Eric, et autres
Publié: (2025)

Stabilizing Policy Optimization via Logits Convexity
par: Chen, Hongzhan, et autres
Publié: (2026)

Less is More: Understanding Word-level Textual Adversarial Attack via n-gram Frequency Descend
par: Lu, Ning, et autres
Publié: (2023)

Oldies but Goldies: The Potential of Character N-grams for Romanian Texts
par: Lupsa, Dana, et autres
Publié: (2025)

The Curious Case of Factual (Mis)Alignment between LLMs' Short- and Long-Form Answers
par: Islam, Saad Obaid ul, et autres
Publié: (2025)

Automatic essay scoring: leveraging Jaccard coefficient and Cosine similaritywith n-gram variation in vector space model approach
par: Cahyani, Andharini Dwi, et autres
Publié: (2025)

How Much Do LLMs Hallucinate across Languages? On Realistic Multilingual Estimation of LLM Hallucination
par: Islam, Saad Obaid ul, et autres
Publié: (2025)

N-gram-like Language Models Predict Reading Time Best
par: Michaelov, James A., et autres
Publié: (2026)

Lossless Acceleration of Large Language Model via Adaptive N-gram Parallel Decoding
par: Ou, Jie, et autres
Publié: (2024)

Memorization and Knowledge Injection in Gated LLMs
par: Pan, Xu, et autres
Publié: (2025)

Towards LLMs Robustness to Changes in Prompt Format Styles
par: Ngweta, Lilian, et autres
Publié: (2025)

Surface Reading LLMs: Synthetic Text and its Styles
par: Bajohr, Hannes
Publié: (2025)