Papers by George Kour
From Zero to Hero: Cold-Start Anomaly Detection (2024.findings-acl)
Copied to clipboard
| Challenge: | Existing anomaly detection methods require previous observations to be effective . contaminated observations are often not observed, making them ineffective . |
| Approach: | They propose a method that adapts a zero-shot anomaly detector to contaminated observations . they propose an evaluation suite consisting of evaluation protocols and metrics . |
| Outcome: | The proposed method adapts the zero-shot anomaly detector to contaminated observations. |
Efficient Agent Evaluation via Diversity-Guided User Simulation (2026.acl-industry)
Copied to clipboard
| Challenge: | Large language models (LLMs) are increasingly deployed as customer-facing agents due to stochastic, multi-turn interactions. |
| Approach: | They propose a snapshot-based, coverage-guided user simulation framework for efficient and systematic exploration of multi-turn agent behavior. |
| Outcome: | Experiments on multi-domain benchmarks show that the proposed method improves failure discovery efficiency and task-level coverage compared to standard linear rollout evaluation without increasing overall cost. |
Balancing via Generation for Multi-Class Text Classification Improvement (2020.findings-emnlp)
Copied to clipboard
| Challenge: | balancing is a known technique for improving classification performance . balancy is based on a balancing policy and a text generation mechanism . |
| Approach: | They propose a balancing-via-generation framework that augments a dataset for more balanced distribution by using a text generation mechanism. |
| Outcome: | The proposed framework can augment a dataset for more balanced distribution while under-sampling others. |
Exploring Straightforward Methods for Automatic Conversational Red-Teaming (2025.naacl-industry)
Copied to clipboard
George Kour, Naama Zwerdling, Marcel Zalmanovici, Ateret Anaby Tavor, Ora Nova Fandina, Eitan Farchi
| Challenge: | Large language models (LLMs) are increasingly used in business dialogue systems but they also pose security and ethical risks. |
| Approach: | They propose to use off-the-shelf large language models to create red-team attacks by eliciting undesired outputs from an attacker LLM. |
| Outcome: | The proposed models can adapt their attack strategies based on prior attempts, but their effectiveness decreases as the alignment of the target model improves. |
Breaking ReAct Agents: Foot-in-the-Door Attack Will Get You In (2025.findings-naacl)
Copied to clipboard
| Challenge: | Indirect prompt injection attacks, prompted by harmless and unrelated requests, can significantly increase the likelihood of the agent performing subsequent malicious actions. |
| Approach: | They propose to implement a simple reflection mechanism that prompts the agent to reassess the safety of its actions during execution, which can help mitigate this vulnerability. |
| Outcome: | The proposed method reduces the success of such attacks by prompting the agent to reassess its actions during execution. |
Effective Red-Teaming of Policy-Adherent Agents (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Model (LLM)-based agents are increasingly used in domains with strict policies, such as refund eligibility or cancellation rules. |
| Approach: | They propose a multi-agent red-teaming system that leverages policy-aware persuasive strategies to undermine a policy-adherence agent in a customer-service scenario. |
| Outcome: | The proposed model outperforms jailbreak methods and tau-break to assess agent's robustness against manipulative user behavior. |
Think Again! The Effect of Test-Time Compute on Preferences, Opinions, and Beliefs of Large Language Models (2025.acl-industry)
Copied to clipboard
| Challenge: | Large Language Models exhibit subjective preferences, opinions, and beliefs, which may shape their behavior, influence advice and recommendations, and potentially reinforce certain viewpoints. |
| Approach: | They developed a benchmark to assess LLMs’ subjective inclinations across societal, cultural, ethical, and personal domains. |
| Outcome: | The proposed benchmark assesses LLMs’ subjective inclinations across societal, cultural, ethical, and personal domains. |