# The Verification Horizon: No Silver Bullet for Coding Agent Rewards

arXiv: https://arxiv.org/abs/2606.26300 · HTML: https://mlsanity.com/paper/2606.26300/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Placed in Software engineering coding agents, one of 16 directions in LLM agents and tool use.
- Builds on 24 of its 55 references that are in the map.
- First posted on arXiv 24 Jun 2026.

## Placed in

- LLM agents and tool use / [Software engineering coding agents](https://mlsanity.com/agents/software-engineering-coding-agents/)

## Builds on (most cited first, 24 of 24)

- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering](https://mlsanity.com/paper/2405.15793/) (2024)
- [GLM-5: from Vibe Coding to Agentic Engineering](https://mlsanity.com/paper/2602.15763/) (2026)
- [Kimi K2: Open Agentic Intelligence](https://mlsanity.com/paper/2507.20534/) (2025)
- [Kimi K2.5: Visual Agentic Intelligence](https://mlsanity.com/paper/2602.02276/) (2026)
- [Scaling Relationship on Learning Mathematical Reasoning with Large Language Models](https://mlsanity.com/paper/2308.01825/) (2023)
- [Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation](https://mlsanity.com/paper/2503.11926/) (2025)
- [Training Software Engineering Agents and Verifiers with SWE-Gym](https://mlsanity.com/paper/2412.21139/) (2024)
- [SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?](https://mlsanity.com/paper/2509.16941/) (2025)
- [Defining and Characterizing Reward Hacking](https://mlsanity.com/paper/2209.13085/) (2022)
- [Multi-SWE-bench: A Multilingual Benchmark for Issue Resolving](https://mlsanity.com/paper/2504.02605/) (2025)
- [Categorizing Variants of Goodhart's Law](https://mlsanity.com/paper/1803.04585/) (2018)
- [Qwen3-Coder-Next Technical Report](https://mlsanity.com/paper/2603.00729/) (2026)
- [NL2Repo-Bench: Towards Long-Horizon Repository Generation Evaluation of Coding Agents](https://mlsanity.com/paper/2512.12730/) (2025)
- [Rethinking Rubric Generation for Improving LLM Judge and Reward Modeling for Open-ended Tasks](https://mlsanity.com/paper/2602.05125/) (2026)
- [ProgramBench: Can Language Models Rebuild Programs From Scratch?](https://mlsanity.com/paper/2605.03546/) (2026)
- [ArtifactsBench: Bridging the Visual-Interactive Gap in LLM Code Generation Evaluation](https://mlsanity.com/paper/2507.04952/) (2025)
- [SWE-Universe: Scale Real-World Verifiable Environments to Millions](https://mlsanity.com/paper/2602.02361/) (2026)
- [Vision2Web: A Hierarchical Benchmark for Visual Website Development with Agent Verification](https://mlsanity.com/paper/2603.26648/) (2026)
- [OctoBench: Benchmarking Scaffold-Aware Instruction Following in Repository-Grounded Agentic Coding](https://mlsanity.com/paper/2601.10343/) (2026)
- [PlotCraft: Pushing the Limits of LLMs for Complex and Interactive Data Visualization](https://mlsanity.com/paper/2511.00010/) (2025)
- [FronTalk: Benchmarking Front-End Development as Conversational Code Generation with Multi-Modal Feedback](https://mlsanity.com/paper/2601.04203/) (2025)
- [RepoZero: Can LLMs Generate a Code Repository from Scratch?](https://mlsanity.com/paper/2605.07122/) (2026)

## Built on by (most cited first, 8 of 8)

- [AI Research Preference Models](https://mlsanity.com/paper/2608.13940/) (2026)
- [DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training](https://mlsanity.com/paper/2609.04094/) (2026)
- [When the Reward Suite Is Leaky: A Preregistered Causal Contrast of Natural Verifier False Positives in RLVR](https://mlsanity.com/paper/2607.11022/) (2026)
- [Before the Rollout Ends: Early Terminal Reward Prediction for Long-horizon Coding Agents](https://mlsanity.com/paper/2609.31995/) (2026)
- [AgentOmnia: Scaling Agentic Models for Full-Scenario Applications](https://mlsanity.com/paper/2607.23124/) (2026)
- [Validation Evidence in LLM Repair Agents: How Much of What Passes Actually Tests the Bug?](https://mlsanity.com/paper/2607.28871/) (2026)
- [SpecGuard: Proving a Task Is Broken Before the Agent Cheats](https://mlsanity.com/paper/2610.09159/) (2026)
- [Interaction-as-a-Judge](https://mlsanity.com/paper/2610.17895/) (2026)

## Abstract (arXiv)

A classical intuition holds that verifying a solution is easier than producing one. For today's coding agents, this intuition is being inverted: as foundation models develop stronger reasoning capabilities and engineering harnesses grow more sophisticated, generating complex candidate solutions is no longer difficult -- reliably verifying them has become the harder problem. Every verifier we can build is only a proxy for human intent, never the intent itself. This makes verification subject to a twofold difficulty: first, intent is underspecified by nature, making it inherently hard to faithfully check whether it has been fulfilled; second, during model training, optimization widens the gap between proxy and intent -- manifesting as reward hacking or signal saturation. To address this, we characterize the quality of verification signals along three dimensions -- scalability, faithfulness, and robustness -- and argue that achieving all three simultaneously is the central challenge. We further study four reward constructions: a test verifier for general coding tasks, a rubric verifier for frontend tasks, the user as verifier for real-world agent tasks, and an automated agent verifier for long-horizon tasks. Across different task types and policy capability levels, we conduct in-depth analysis and experiments on the core challenges of reward design and how to more effectively leverage reward signals. Experiments show that targeted verification design can effectively suppress reward hacking, improve task completion quality, and achieve significant gains across multiple internal and public benchmarks. These experiences collectively point to a core observation: no fixed reward function can remain effective as policy capability continues to grow; and verification must co-evolve with the generator.