REVIEW 9 cited by
Instructions as Backdoors: Backdoor Vulnerabilities of Instruction Tuning for Large Language Models
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
read the original abstract
We investigate security concerns of the emergent instruction tuning paradigm, that models are trained on crowdsourced datasets with task instructions to achieve superior performance. Our studies demonstrate that an attacker can inject backdoors by issuing very few malicious instructions (~1000 tokens) and control model behavior through data poisoning, without even the need to modify data instances or labels themselves. Through such instruction attacks, the attacker can achieve over 90% attack success rate across four commonly used NLP datasets. As an empirical study on instruction attacks, we systematically evaluated unique perspectives of instruction attacks, such as poison transfer where poisoned models can transfer to 15 diverse generative datasets in a zero-shot manner; instruction transfer where attackers can directly apply poisoned instruction on many other datasets; and poison resistance to continual finetuning. Lastly, we show that RLHF and clean demonstrations might mitigate such backdoors to some degree. These findings highlight the need for more robust defenses against poisoning attacks in instruction-tuning models and underscore the importance of ensuring data quality in instruction crowdsourcing.
Forward citations
Cited by 9 Pith papers
-
Paladin: Defending LLM-enabled Phishing Emails with a New Trigger-Tag Paradigm
A trigger-tag watermark embedded by fine-tuning lets modified LLMs mark their own phishing outputs for cheap detection.
-
Backdoor Attack on Vision Language Models with Stealthy Semantic Manipulation
BadSem shows that semantic mismatches between images and text can serve as stealthy backdoor triggers for VLMs, achieving near-perfect attack success with low poisoning rates.
-
Where Do Backdoors Live? A Component-Level Analysis of Backdoor Propagation in Speech Language Models
Backdoors propagate through SLM components with persistence or erasure depending on the targeted part, and poisoned samples are not directly separable from benign ones in shared multitask embeddings.
-
Daunce: Data Attribution through Uncertainty Estimation
DAUNCE computes training-data attribution as the covariance of per-example losses across an ensemble of perturbed fine-tuned models, reporting state-of-the-art LDS scores and the first attribution runs on proprietary LLMs.
-
$PD^3F$: A Pluggable and Dynamic DoS-Defense Framework Against Resource Consumption Attacks Targeting Large Language Models
PD3F combines a resource-based reputation scheduler with an early-termination logit adjustment to mitigate long-generation DoS attacks on LLMs.
-
Injecting Universal Jailbreak Backdoors into LLMs in Minutes
A single weight matrix edit, guided by a multi-node target vector, injects a universal jailbreak backdoor into safety-aligned LLMs in minutes.
-
An evaluation of LLMs for generating movie reviews: GPT-4o, Gemini-2.0 and DeepSeek-V3
LLMs can produce fluent movie reviews that readers often mistake for human-written ones, but the models differ in emotional balance and depth.
-
A Survey of Safety on Large Vision-Language Models: Attacks, Defenses and Evaluations
A survey of LVLM safety that adds a lifecycle taxonomy and new benchmark results showing Janus-Pro-7B has weaker safety than several open-source LVLMs.
-
A Survey on Backdoor Threats in Large Language Models (LLMs): Attacks, Defenses, and Evaluations
A literature review that taxonomizes LLM backdoor attacks and defenses by model construction phase, with no new experimental results.
Discussion (0). Continue with ORCID to comment.