Prompt injection attacks can self-replicate across LLM agents in multi-agent systems, enabling data theft, misinformation, and system disruption while propagating silently.
hub Canonical reference
Instruction Tuning with GPT-4
Canonical reference. 94% of citing Pith papers cite this work as background.
65
Pith papers citing it
Background
94% of classified citations
abstract
Prior work has shown that finetuning large language models (LLMs) using machine-generated instruction-following data enables such models to achieve remarkable zero-shot capabilities on new tasks, and no human-written instructions are needed. In this paper, we present the first attempt to use GPT-4 to generate instruction-following data for LLM finetuning. Our early experiments on instruction-tuned LLaMA models show that the 52K English and Chinese instruction-following data generated by GPT-4 leads to superior zero-shot performance on new tasks to the instruction-following data generated by previous state-of-the-art models. We also collect feedback and comparison data from GPT-4 to enable a comprehensive evaluation and reward model training. We make our data generated using GPT-4 as well as our codebase publicly available.
hub tools
citation-role summary
background 17
dataset 1
citation-polarity summary
representative citing papers
SpliceLeak is the first end-to-end side-channel attack on non-prefix KV cache in RAG, using Step-Wave timing leaks to fingerprint private prompt lengths and extract tokens with up to 100% success using 63 requests per token on vLLM+LMCache.
FloatDoor uses two LoRA adapters to create the first input-independent backdoor that triggers adversary-chosen behavior only on a target platform while remaining benign elsewhere.
HARP is a train-based data selector for LLM finetuning that uses hierarchical active region pruning and empirical Bayes posteriors to achieve up to 8.9 point gains with roughly 7 times fewer training examples.
D³ introduces a dynamic directional graph-constrained framework that models sample interactions via loss dependencies to derive an optimized training sequence for LLMs.
The paper introduces an outlier-injection attack that induces targeted weight collapse in LLMs under advanced quantization schemes including AWQ, GPTQ, and GGUF I-quants.
StepCodeReasoner aligns code reasoning with verifiable stepwise execution traces via print anchors and bi-level GRPO reinforcement learning, reaching SOTA results on CRUXEval (91.1%) and LiveCodeBench (86.5%) for a 7B model.
RouteHijack is a routing-aware jailbreak that identifies safety-critical experts via activation contrast and optimizes suffixes to suppress them, reaching 69.3% average attack success rate on seven MoE LLMs with strong transfer to variants and VLMs.
ProjLens shows that backdoor parameters in MLLMs are encoded in low-rank subspaces of the projector and that embeddings shift toward the target direction with magnitude linear in input norm, activating only on poisoned samples.
A new auditing framework reveals widespread behavioral entanglement among LLMs and shows that reweighting ensembles based on measured independence improves verification accuracy by up to 4.5%.
MIDUS replaces duplicated FFN branches in depth up-scaling with head-wise memory layers using product-key retrieval and HIVE to deliver lightweight, head-conditioned residual capacity.
Prefix-RFT blends SFT and RFT via prefix sampling from demonstrations to outperform standalone SFT, RFT, and mixed-policy baselines on math reasoning problems.
Iterative self-rewarding via LLM-as-Judge in DPO training on Llama 2 70B improves instruction following and self-evaluation, outperforming GPT-4 on AlpacaEval 2.0.
HallusionBench shows GPT-4V reaches only 31.42% accuracy on paired questions testing language hallucination and visual illusion in LVLMs, with other models below 16%.
QLoRA finetunes 4-bit quantized LLMs via LoRA adapters to match full-precision performance while using far less memory, enabling 65B-scale training on single GPUs and producing Guanaco models near ChatGPT level.
LLaVA is trained on GPT-4 generated visual instruction data to achieve 85.1% relative performance to GPT-4 on synthetic multimodal tasks and 92.53% accuracy on Science QA.
LLaMA-Adapter turns frozen LLaMA 7B into a capable instruction follower using only 1.2M new parameters and zero-init attention, matching Alpaca while extending to image-conditioned reasoning on ScienceQA and COCO.
QuantGuard is a pre-quantization method using differentiable rounding controls, error-guided reversal constraints, output consistency, and weight regularization on a small calibration set to suppress quantization-conditioned backdoors while preserving performance.
TOPS formulates visual token pruning as constructing Token Optimal Preservation Sets using three information-theoretic principles and demonstrates superior performance on MLLM benchmarks.
OPPO applies RL with an Omni-Perception Reward and masked-input KL loss to boost cue utilization and suppress hallucinations in emotion reasoning MLLMs, claiming SOTA results on MER-UniBench, MME-Emotion, and MEP-Bench.
Brick-DICL applies dynamic in-context learning with two RAG stages and multi-LLM filtering to automate mapping of BMS points to the 936-class Brick ontology, claiming accuracy gains and reduced manual verification.
SafeSteer restricts reverse KL penalty to safety tokens selected via activation steering, achieving strong safety on seven benchmarks with minimal degradation on five capability benchmarks using only 100 harmful samples and no general data.
A dataset-agnostic framework converts text tool-calling benchmarks to paired audio evaluations via TTS, speaker variation and noise, then evaluates seven omni-modal models showing model- and task-dependent performance with small text-to-voice gaps.
Diagnoses mask prior drift and positional attention collapse in LDVLMs and introduces two plug-and-play decoding interventions that raise long-form generation quality without retraining.
citing papers explorer
-
How Independent are Large Language Models? A Statistical Framework for Auditing Behavioral Entanglement and Reweighting Verifier Ensembles
A new auditing framework reveals widespread behavioral entanglement among LLMs and shows that reweighting ensembles based on measured independence improves verification accuracy by up to 4.5%.
-
TOPS: First-Principles Visual Token Pruning via Constructing Token Optimal Preservation Sets for Efficient MLLM Inference
TOPS formulates visual token pruning as constructing Token Optimal Preservation Sets using three information-theoretic principles and demonstrates superior performance on MLLM benchmarks.
-
Omni-Perception Policy Optimization for Multimodal Emotion Reasoning
OPPO applies RL with an Omni-Perception Reward and masked-input KL loss to boost cue utilization and suppress hallucinations in emotion reasoning MLLMs, claiming SOTA results on MER-UniBench, MME-Emotion, and MEP-Bench.
-
Brick-DICL: Dynamic In-Context Learning for Automated Brick Schema Classification
Brick-DICL applies dynamic in-context learning with two RAG stages and multi-LLM filtering to automate mapping of BMS points to the 936-class Brick ontology, claiming accuracy gains and reduced manual verification.
-
SafeSteer: Localized On-Policy Distillation for Efficient Safety Alignment
SafeSteer restricts reverse KL penalty to safety tokens selected via activation steering, achieving strong safety on seven benchmarks with minimal degradation on five capability benchmarks using only 100 harmful samples and no general data.
-
LLM-X: A Scalable Negotiation-Oriented Exchange for Communication Among Personal LLM Agents
LLM-X is a scalable architecture for direct negotiation and communication among personal LLM agents, featuring federated gateways, typed protocols, and policy enforcement, shown stable in experiments with up to 12 agents.
-
Generalization in LLM Problem Solving: The Case of the Shortest Path
LLMs show strong spatial generalization to unseen maps in shortest-path tasks but fail length scaling due to recursive instability, with data coverage setting hard limits.
-
Improving Collaborative Storytelling with a Multi-Agent Framework Based on Large Language Models
An iterative writer-editor multi-agent LLM process improves perceived story quality in simulations of child collaborative storytelling.
-
Contrastive Reasoning Alignment: Reinforcement Learning from Hidden Representations
CRAFT uses contrastive representation learning and RL on hidden states to align reasoning models for improved safety against jailbreaks, reporting 79% and 87.7% gains over base models.