Kerkoporta: Injecting Structured-Output Backdoors in Reasoning Coding Agents
cs.CR, cs.AI
Submitted: 2026-10-29
Updated: 2026-10-29
Code: https://github.com/NovaSky-AI/SkyRL
Terminology
Sources
- Kimi K3: Open Frontier Intelligence
- DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression
- On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training Stability
- Open Problems in Mechanistic Interpretability
- Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
- MetaBackdoor: Exploiting Positional Encoding as a Backdoor Attack Surface in LLMs
- ReAct: Synergizing Reasoning and Acting in Language Models
- Poisoning Attacks against Support Vector Machines
- Manipulating Machine Learning: Poisoning Attacks and Countermeasures for Regression Learning
- Can You Really Backdoor Federated Learning?
- Tmax: A simple recipe for terminal agents
- RL's Razor: Why Online Reinforcement Learning Forgets Less
- Retaining by Doing: The Role of On-Policy Data in Mitigating Forgetting
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- LoRA: Low-Rank Adaptation of Large Language Models
- Evading Chain-of-Thought Monitoring Through Model Poisoning
- Backdoor Attacks and Countermeasures in Natural Language Processing Models: A Comprehensive Security Review
- A Survey on Backdoor Threats in Large Language Models (LLMs): Attacks, Defenses, and Evaluations
- Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samples
Related papers
- From Strategy Search to Policy Learning: MCTS-Guided Adversarial Co-Evolution for LLM Safety Alignment
- Privacy Leakage from Recurrent State in LLM Inference
- Incrementally Verifiable Oracle Aided Computation
- Privilege and confidentiality in generative AI workflows
- BadMoE: Backdooring Mixture-of-Experts LLMs via Routing-Aware Triggers and Task-Underutilized Experts
- CoVault: Secure, Scalable Analytics of Personal Data