LatentSift: Policy-State Filtering for Token-Efficient Verification of Software Engineering Agents
cs.SE, cs.AI
Submitted: 2026-09-28
Updated: 2026-09-28
Terminology
Sources
- Process Reward Models for LLM Agents: Practical Framework and Directions
- Training Verifiers to Solve Math Word Problems
- SWE-Shepherd: Advancing PRMs for Reinforcing Code Agents
- CodeMonkeys: Scaling Test-Time Compute for Software Engineering
- CWM: An Open-Weights LLM for Research on Code Generation with World Models
- Can LLMs Predict Their Own Failures? Self-Awareness via Internal Circuits
- SWE-TRACE: Optimizing Long-Horizon SWE Agents Through Rubric Process Reward Models and Heuristic Test-Time Scaling
- Language Models (Mostly) Know What They Know
- Semantic Entropy Probes: Robust and Cheap Hallucination Detection in LLMs
- Let's reward step by step: Step-Level reward model as the Navigators for Reasoning
- Solving math word problems with process- and outcome-based feedback
- Scalable Supervision for Software Agents via Patch Reasoning
- Reasoning Models Know When They're Right: Probing Hidden States for Self-Verification
- Parallel-Probe: Towards Efficient Parallel Thinking via 2D Probing
- LLMs Improving LLMs: Agentic Discovery for Test-Time Scaling
- Scaling Test-time Compute for LLM Agents
Related papers
- Falsification-Based Verification of LLM-Generated Optimization Models: Sound Test Batteries and Their Detection Limits
- GitSkills: A Dataset of Agent Skills on GitHub
- SABER: Benchmarking Operational Safety of LLM Coding Agents in Stateful Project Workspaces
- PackMonitor: Enabling Zero Package Hallucinations Through Decoding-Time Monitoring
- IntentCoding: Amplifying User Intent in Code Generation
- Incentives and Outcomes in Bug Bounties