Pith. sign in

REVIEW 8 cited by

Persistent Pre-Training Poisoning of LLMs

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2410.13722 v1 pith:DC564HSP submitted 2024-10-17 cs.CR cs.AI

Persistent Pre-Training Poisoning of LLMs

classification cs.CR cs.AI
keywords poisoningpre-trainingmodelsattacksdatasetslanguagedenial-of-servicefour
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Large language models are pre-trained on uncurated text datasets consisting of trillions of tokens scraped from the Web. Prior work has shown that: (1) web-scraped pre-training datasets can be practically poisoned by malicious actors; and (2) adversaries can compromise language models after poisoning fine-tuning datasets. Our work evaluates for the first time whether language models can also be compromised during pre-training, with a focus on the persistence of pre-training attacks after models are fine-tuned as helpful and harmless chatbots (i.e., after SFT and DPO). We pre-train a series of LLMs from scratch to measure the impact of a potential poisoning adversary under four different attack objectives (denial-of-service, belief manipulation, jailbreaking, and prompt stealing), and across a wide range of model sizes (from 600M to 7B). Our main result is that poisoning only 0.1% of a model's pre-training dataset is sufficient for three out of four attacks to measurably persist through post-training. Moreover, simple attacks like denial-of-service persist through post-training with a poisoning rate of only 0.001%.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 8 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Safety from Honesty in a Disinterested AI Predictor

    cs.AI 2026-06 conditional novelty 7.0

    Under consequence-invariant posterior training and sparsity of coordinated harm patterns, the training mass on dangerous guarded Predictors is bounded by C_bad times R_shell.

  2. Loss Landscape Poisoning: Targeted Extraction of Unseen Training Data from LLMs

    cs.CR 2026-06 unverdicted novelty 7.0

    Poisoning training data reshapes the loss landscape to enable targeted extraction of unseen data from LLMs with high success rates in language and vision-language models.

  3. RogueMerge: Robust and Unified Attacks against LLM Model Merging

    cs.CR 2026-06 unverdicted novelty 7.0

    RogueMerge is a unified attack method that jointly optimizes task vectors to succeed after merging, using stochastic min-max simulation for unknown merging settings and a Taylor-approximated DRO for prompt generalizat...

  4. Security--Fidelity Tradeoffs: The Hidden Cost of Prompt Injection Defense

    cs.CR 2026-06 unverdicted novelty 6.0

    Prompt injection defenses create a security-fidelity tradeoff with no model or defense achieving both high security and high fidelity on the SecFid benchmark across 1,168 examples.

  5. Safety from Honesty in a Disinterested AI Predictor

    cs.AI 2026-06 unverdicted novelty 6.0

    A disinterested Bayesian Predictor trained on contextualized statements has low probability of producing harmful agency because dangerous behaviors require rare coordinated underestimation of harm with no training sig...

  6. Iterative Finetuning is Mostly Idempotent

    cs.AI 2026-05 unverdicted novelty 6.0

    Iterative self-finetuning of LLMs mostly fails to amplify seeded behavioral traits, with amplification limited to specific DPO setups and often harming coherence.

  7. When AI reviews science: Can we trust the referee?

    cs.AI 2026-04 unverdicted novelty 6.0

    AI peer review systems are vulnerable to prompt injections, prestige biases, assertion strength effects, and contextual poisoning, as demonstrated by a new attack taxonomy and causal experiments on real conference sub...

  8. LLM Hypnosis: Exploiting User Feedback for Unauthorized Knowledge Injection to All Users

    cs.CL 2025-07 unverdicted novelty 6.0

    A single attacker can use strategic upvoting and downvoting on language model outputs to inject facts, security flaws, or fake news that persist in the model for all users after preference tuning.