Image de couverture de livre

AnomaLLMy -- Detecting anomalous tokens in black-box LLMs through low-confidence single-token predictions

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteur principal:	Witold, Waligóra
Format:	Preprint
Publié:	2024
Sujets:	Computation and Language Artificial Intelligence
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

The pitfalls of next-token prediction
par: Bachmann, Gregor, et autres
Publié: (2024)

Prediction hubs are context-informed frequent tokens in LLMs
par: Nielsen, Beatrix M. G., et autres
Publié: (2025)

Jacobian Scopes: token-level causal attributions in LLMs
par: Liu, Toni J. B., et autres
Publié: (2026)

Language models are better than humans at next-token prediction
par: Shlegeris, Buck, et autres
Publié: (2022)

Looking beyond the next token
par: Thankaraj, Abitha, et autres
Publié: (2025)

Towards Nepali-language LLMs: Efficient GPT training with a Nepali BPE tokenizer
par: Shrestha, Adarsha, et autres
Publié: (2025)

Shaping capabilities with token-level data filtering
par: Rathi, Neil, et autres
Publié: (2026)

You only need 4 extra tokens: Synergistic Test-time Adaptation for LLMs
par: Xu, Yijie, et autres
Publié: (2025)

On the token distance modeling ability of higher RoPE attention dimension
par: Hong, Xiangyu, et autres
Publié: (2024)

Scaling Transformer to 1M tokens and beyond with RMT
par: Bulatov, Aydar, et autres
Publié: (2023)

Visualizing token importance for black-box language models
par: Rauba, Paulius, et autres
Publié: (2025)

Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context
par: Gemini Team, et autres
Publié: (2024)

Scaled and Inter-token Relation Enhanced Transformer for Sample-restricted Residential NILM
par: Rahman, Minhajur, et autres
Publié: (2024)

Roll the dice & look before you leap: Going beyond the creative limits of next-token prediction
par: Nagarajan, Vaishnavh, et autres
Publié: (2025)

Interpretable Next-token Prediction via the Generalized Induction Head
par: Kim, Eunji, et autres
Publié: (2024)

Unused information in token probability distribution of generative LLM: improving LLM reading comprehension through calculation of expected values
par: Zawistowski, Krystian
Publié: (2024)

DeepMLF: Multimodal language model with learnable tokens for deep fusion in sentiment analysis
par: Georgiou, Efthymios, et autres
Publié: (2025)

SubTokenTest: A Practical Benchmark for Real-World Sub-token Understanding
par: Hou, Shuyang, et autres
Publié: (2026)

Surrogate modeling for interpreting black-box LLMs in medical predictions
par: Han, Changho, et autres
Publié: (2026)

Large-scale cloze evaluation reveals that token prediction tasks are neither lexically nor semantically aligned
par: Jacobs, Cassandra L., et autres
Publié: (2024)

COMPACT: Common-token Optimized Model Pruning Across Channels and Tokens
par: Kwek, Eugene, et autres
Publié: (2025)

Not all tokens are created equal: Perplexity Attention Weighted Networks for AI generated text detection
par: Miralles-González, Pablo, et autres
Publié: (2025)

All or None: Identifiable Linear Properties of Next-token Predictors in Language Modeling
par: Marconato, Emanuele, et autres
Publié: (2024)

Essential-Web v1.0: 24T tokens of organized web data
par: AI, Essential, et autres
Publié: (2025)

Boundless Byte Pair Encoding: Breaking the Pre-tokenization Barrier
par: Schmidt, Craig W., et autres
Publié: (2025)

Local Explanations and Self-Explanations for Assessing Faithfulness in black-box LLMs
par: Fragkathoulas, Christos, et autres
Publié: (2024)

Is Sanskrit the most token-efficient language? A quantitative study using GPT, Gemini, and SentencePiece
par: Kumar, Anshul
Publié: (2026)

Not a nuisance but a useful heuristic: Outlier dimensions favor frequent tokens in language models
par: Macocco, Iuri, et autres
Publié: (2025)

Finetuning LLMs for EvaCun 2025 token prediction shared task
par: Jon, Josef, et autres
Publié: (2025)

Explaining multimodal LLMs via intra-modal token interactions
par: Liang, Jiawei, et autres
Publié: (2025)

Automated Detection of Pre-training Text in Black-box LLMs
par: Hu, Ruihan, et autres
Publié: (2025)

Why do LLMs attend to the first token?
par: Barbero, Federico, et autres
Publié: (2025)

Not all tokens are needed(NAT): token efficient reinforcement learning
par: Sang, Hejian, et autres
Publié: (2026)

On multi-token prediction for efficient LLM inference
par: Mehra, Somesh, et autres
Publié: (2025)

MALicious INTent Dataset and Inoculating LLMs for Enhanced Disinformation Detection
par: Modzelewski, Arkadiusz, et autres
Publié: (2026)

Interpreting token compositionality in LLMs: A robustness analysis
par: Aljaafari, Nura, et autres
Publié: (2024)

Scaling FP8 training to trillion-token LLMs
par: Fishman, Maxim, et autres
Publié: (2024)

Tokenization counts: the impact of tokenization on arithmetic in frontier LLMs
par: Singh, Aaditya K., et autres
Publié: (2024)

Augmenting Black-box LLMs with Medical Textbooks for Biomedical Question Answering
par: Wang, Yubo, et autres
Publié: (2023)

On the attribution of confidence to large language models
par: Keeling, Geoff, et autres
Publié: (2024)