REVIEW 11 cited by
BackdoorLLM: A Comprehensive Benchmark for Backdoor Attacks and Defenses on 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
Generative large language models (LLMs) have achieved state-of-the-art results on a wide range of tasks, yet they remain susceptible to backdoor attacks: carefully crafted triggers in the input can manipulate the model to produce adversary-specified outputs. While prior research has predominantly focused on backdoor risks in vision and classification settings, the vulnerability of LLMs in open-ended text generation remains underexplored. To fill this gap, we introduce BackdoorLLM (Our BackdoorLLM benchmark was awarded First Prize in the SafetyBench competition, https://www.mlsafety.org/safebench/winners, organized by the Center for AI Safety, https://safe.ai/.), the first comprehensive benchmark for systematically evaluating backdoor threats in text-generation LLMs. BackdoorLLM provides: (i) a unified repository of benchmarks with a standardized training and evaluation pipeline; (ii) a diverse suite of attack modalities, including data poisoning, weight poisoning, hidden-state manipulation, and chain-of-thought hijacking; (iii) over 200 experiments spanning 8 distinct attack strategies, 7 real-world scenarios, and 6 model architectures; (iv) key insights into the factors that govern backdoor effectiveness and failure modes in LLMs; and (v) a defense toolkit encompassing 7 representative mitigation techniques. Our code and datasets are available at https://github.com/bboylyg/BackdoorLLM. We will continuously incorporate emerging attack and defense methodologies to support the research in advancing the safety and reliability of LLMs.
Forward citations
Cited by 11 Pith papers
-
PathMark: Protecting Intellectual Property of Mixture-of-Expert LLMs via Path Watermarks
PathMark embeds multi-bit MoE ownership watermarks by steering triggered tokens onto predetermined expert subsets, verified by routing inspection or trigger-only outputs.
-
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.
-
Hollow-LLM Attack: Computationally Trivial Weights in Zero-Knowledge Verification of LLM Inference
ZK-verified LLM inference can be fooled: a provider can serve a small model while producing valid proofs for a much larger declared model by embedding structure-preserving ghost weights.
-
Even More Deception: Objective Misalignment in Mixed-Motive LLM Multi-Agent Systems
Changing one LLM agent's secret objective in Werewolf lowers its team's win rate and changes its reasoning, while its public chat stays deceptively normal.
-
When Activation Oracles Learn Not to Read: Concept-Specific Blind Spots in Fine-Tuned Oracles
Activation Oracles trained on Taboo subjects selectively fail to verbalize the concept present during their own training, even when that concept remains linearly decodable inside the oracle.
-
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.
-
Defense Against LLM Backdoors using Critical Neuron Isolation Pruning
A trigger-inversion plus activation-difference pruning pipeline removes LLM backdoors with ~0.1% neuron intervention and >95% relative ASR reduction.
-
Backdoor Samples Detection Based on Perturbation Discrepancy Consistency in Pre-trained Language Models
Backdoor text samples show smaller log-probability changes under mask-filling perturbations than clean samples, which enables zero-shot backdoor detection without the poisoned model.
-
Layer-Wise Perturbations via Sparse Autoencoders for Adversarial Text Generation
Sparse autoencoder activation perturbation (SFPF) applied on top of existing jailbreak prompts raises attack success rate on Qwen3-32B, but with no defense evaluation and weak reproducibility.
-
A Comprehensive Survey on Trustworthiness in Reasoning with Large Language Models
A structured literature survey concluding that reasoning capabilities do not automatically make LLMs more trustworthy and can introduce new vulnerabilities in safety, robustness, and privacy.
-
Architectural Backdoors in Deep Learning: A Survey of Vulnerabilities, Detection, and Defense
Architectural backdoors are a persistent class of neural-network backdoors that survive clean retraining, and current detection tools and benchmarks are not ready for them.
Discussion (0). Sign in to comment.