Scalable Weak-to-Strong Verifier Hardening for Agent Environments
cs.CR, cs.AI, cs.LG, cs.MA
Submitted: 2026-10-11
Updated: 2026-10-11
Comments: Project page: https://verifier-hardening.github.io/
Code: https://github.com/SWE-bench/SWE-bench
Project page: https://verifier-hardening.github.io
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Kevin: Multi-Turn RL for Generating CUDA Kernels
- Let It Flow: Agentic Crafting on Rock and Roll, Building the ROME Model within an Open Agentic Learning Ecosystem
- SETA: Scaling Environments for Terminal Agents
- EvilGenie: A Reward Hacking Benchmark
- Reward Hacking Benchmark: Measuring Exploits in LLM Agents with Tool Use
- CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation
- Towards Robust Agentic CUDA Kernel Benchmarking, Verification, and Optimization
- Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision
- Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models
- School of Reward Hacks: Hacking harmless tasks generalizes to misaligned behavior in LLMs
- ImpossibleBench: Measuring LLMs' Propensity of Exploiting Test Cases
- SpecBench: Measuring Reward Hacking in Long-Horizon Coding Agents
- Hodoscope: Unsupervised Monitoring for AI Misbehaviors
- Detecting Safety Violations Across Many Agent Traces
- RvB: Automating AI System Hardening via Iterative Red-Blue Games
- Proactive Hardening of LLM Defenses with HASTE
- Mutation-Guided LLM-based Test Generation at Meta
- AI safety via debate
- Prover-Verifier Games improve legibility of LLM outputs
- GAR: Generative Adversarial Reinforcement Learning for Formal Theorem Proving
Related papers
- SoK: AI-Augmented Binary Reversing
- Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
- Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
- Efficient Fuzzy PSI under One-Sided Assumptions
- Sealing the Audit-Runtime Gap for LLM Skills
- Token Composition: A Graph Based on EVM Logs