# Beyond pass@1: A Reliability Science Framework for Long-Horizon LLM Agents

arXiv: https://arxiv.org/abs/2603.29231 · HTML: https://mlsanity.com/paper/2603.29231/

## Key facts

- Cited by 10 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 3 of its 5 references that are in the map.
- First posted on arXiv 31 Mar 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, 3 of 3)

- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [WebArena: A Realistic Web Environment for Building Autonomous Agents](https://mlsanity.com/paper/2307.13854/) (2023)
- [τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains](https://mlsanity.com/paper/2406.12045/) (2024)

## Built on by (most cited first, 10 of 10)

- [Why Retrying Fails: Context Contamination in LLM Agent Pipelines](https://mlsanity.com/paper/2605.08563/) (2026)
- [Measuring the Unmeasurable: Markov Chain Reliability for LLM Agents](https://mlsanity.com/paper/2604.24579/) (2026)
- [Relational Intervention During Functional Collapse in Large Language Models: A Lexical-Statistical Ablation and a Structure x Register Factorial](https://mlsanity.com/paper/2606.00935/) (2026)
- [No Task Fails Every Time: Why One-Shot Audits Are Structurally Blind to Agent Damage](https://mlsanity.com/paper/2608.15286/) (2026)
- [What Does Context Compression Cost an Agent? Interaction Costs Unrevealed by Task-Completion Metrics](https://mlsanity.com/paper/2608.16370/) (2026)
- [Selection Is Retrieval, Abstention Is Not: On-Device Tool Routing over 70 Korean-English Actions](https://mlsanity.com/paper/2609.18672/) (2026)
- [When the Agent Becomes the Kernel: A Systematization of Security on the Path to AI-Native Operating Systems](https://mlsanity.com/paper/2609.23700/) (2026)
- [Agents Are Systems, Not Models: Rethinking Agentic Evaluation](https://mlsanity.com/paper/2610.01618/) (2026)
- [OpenGameEval: Benchmarking Agentic Programming and Exploration in a Stateful Game Engine](https://mlsanity.com/paper/2610.02563/) (2026)
- [When Agent Evaluation Becomes Unreliable: An Empirical Study of Oracle Failures and Configuration-Dependent Measurement in an LLM Agent Benchmark](https://mlsanity.com/paper/2610.16067/) (2026)

## Abstract (arXiv)

Existing benchmarks measure capability -- whether a model succeeds on a single attempt -- but production deployments require reliability -- consistent success across repeated attempts on tasks of varying duration. We show these properties diverge systematically as task duration grows, and that pass@1 on short tasks is structurally blind to this divergence. We introduce a reliability science framework for long-horizon LLM agents with four metrics: Reliability Decay Curve (RDC), Variance Amplification Factor (VAF), Graceful Degradation Score (GDS), and Meltdown Onset Point (MOP). We evaluate 10 models across 23,392 episodes on a 396-task benchmark spanning four duration buckets and three domains. Key findings: (1) reliability decay is domain-stratified -- SE GDS drops from 0.90 to 0.44 while document processing is nearly flat (0.74 to 0.71); (2) VAF bifurcates by capability tier -- high VAF is a capability signature, not an instability signal; (3) capability and reliability rankings diverge substantially, with multi-rank inversions at long horizons; (4) frontier models have the highest meltdown rates (up to 19%) because they attempt ambitious multi-step strategies that sometimes spiral; and (5) memory scaffolds universally hurt long-horizon performance across all 10 models. These results motivate reliability as a first-class evaluation dimension alongside capability.