Skip to content
Preprint

FragToken: Amplifying LLM Inference Costs through Noncanonical Token Generation

Sep 2026 · 0 citations · 38 references
Computer Science

TL;DR

FragToken is proposed, a training-time framework that combines source-model self-distillation, capacity-aware filtering and budgeting, and BPE-Aligned Merging to induce fragmented generation under ordinary prompts while largely preserving model utility.

Abstract

As large language model (LLM) inference becomes increasingly expensive, resource-consumption attacks pose a growing threat to model providers. Existing attacks typically amplify cost by inducing abnormally long or repetitive outputs on attacker-controlled or triggered requests, making them easier to detect and limiting their deployment-wide impact when benign traffic dominates. In this work, we uncover a previously overlooked token-level attack surface arising from the many-to-one mapping from token sequences to decoded text. Although standard LLMs predominantly generate the canonical token sequences induced by their tokenizers, the same text can also be represented by substantially longer non-canonical sequences. This representational flexibility exposes a new avenue for resource-consumption attacks: an attacker can train the model to favor such sequences, systematically increasing the number of autoregressive decoding steps without a proportional increase in visible response length. However, we empirically find that directly maximizing token fragmentation substantially degrades model utility, producing conspicuous answer-quality failures that undermine attack stealthiness. To address this challenge, we propose FragToken, a training-time framework that combines source-model self-distillation, capacity-aware filtering and budgeting, and BPE-Aligned Merging to induce fragmented generation under ordinary prompts while largely preserving model utility. We evaluate FragToken on four LLMs across three benchmarks. Across the four models, FragToken achieves a three-benchmark average token inflation ratio (TIR) ranging from 1.99 to 2.46, while causing only minor degradation in model utility. Our work reveals a covert LLM supply-chain threat that increases inference cost without requiring large volumes of attack requests while largely preserving utility.

View source

Similar papers

Controlled Decoding Attacks on Black-Box LLMs

This work introduces \method{}, a framework for jailbreaking through text-only continuation interfaces that permit repeated sampling and assistant-prefix continuation, and achieves the highest mean score most comparisons against baselines.

Jesson Wang, Shawn Li, Wei Yang et al. · 2 citations
#artificial intelligence Preprint Oct 2026

Exploring Jailbreak Vulnerability Across Depth-Based Dynamic Inference in Large Language Models

This paper makes the following key observations: Shallow exits are consistently more robust to jailbreak attacks than deeper ones; Exit depth affects how easily the adversarial suffix can be optimized, with deeper exits reaching lower attack loss; and adversarial suffixes affect utility uniformly across exit depths.

Lass Najmuldeen, Seyedhamidreza Mousavi · 0 citations
#natural language process... Preprint Oct 2026

Reward Stealing Attack on Large Language Models

Experiments demonstrate that a single recovered reward generalizes across prompts and diverse models to reveal a fundamental alignment vulnerability, enabling ReSA to significantly outperform existing attacks in effectiveness and transferability.

Jia-Ming Qian, Peng-Yang Zhou, Jia-He Xu et al. · 0 citations
#artificial intelligence Preprint Sep 2026

Distillation Defenses Easily Break After Reinforcement Learning

It is argued that a more realistic threat model includes further training with reinforcement learning after distillation, and it is shown that simple attacks can steal reasoning capabilities from existing closed-source language models using data easily obtainable from current APIs, yielding reasoning improvements equiv...

S. Javaheri, Alexander Panfilov, O. Britton et al. · 0 citations
#natural language process... Preprint Oct 2026

HeadEdit: Calibrating Language Model Behavior Through the Frozen Unembedding Matrix

Alignment does not eliminate behavioral errors in language models. Models may still refuse benign requests, call unnecessary tools, or yield to false user claims. Current methods mitigate such errors as a computation problem, and rarely explore if the desired behavior is already encoded in the model's representation. M...

Zirui He, Hai-Yan Zhao, Jing-Yu Hu et al. · 0 citations
#artificial intelligence Preprint Oct 2026

Speedbumps: Rejection Attacks on Speculative Decoding

This work introduces two attacks which append an adversarial suffix to attacker-controlled content to degrade speculative decoding on a victim's prompts, identifying the draft-target interaction of speculative decoding as a realistic attack surface through which adversarial inputs can inflate inference costs.

Adam Y. J. Jones, Yuan Yuan, Sergio Maffeis · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.