Rebuilding problem parts shapes language agent fault recovery results
Re-derivability Decides What a Staged Agent Pipeline Recovers After an Upstream Fault
Artificial IntelligenceComputation and LanguageMachine Learning
Summary
This paper looks at how a chain of AI language agents handle mistakes that happen early in their process. The key factor is "re-derivability," which means how much a later agent can rebuild what it needs from the original task despite earlier faults. The authors found that grounding inspectors in the original problem helps fix errors better than blind inspection. Also, repairing faults early in the chain brings the biggest improvements, while later fixes add little benefit. This shows that the ability to reconstruct information strongly affects how well multi-agent pipelines recover from mistakes.
What this means in practice
- •For ai system engineers: Design multi-agent AI pipelines that re-derive problem aspects early to improve fault recovery efficiency and accuracy.
- •For natural language processing developers: Enhance language model chains by grounding inspection stages in the original problem to better detect and mitigate upstream faults.
Authors
Tianqi Bu, YuXuan Peng, Junteng Tu, Henghui Xiao
Abstract
One variable sets what an upstream fault costs a staged pipeline of language-model agents: re-derivability, how much of what a stage needs it can rebuild from the original problem. Grounding an inspector agent in that problem is worth +0.608 [+0.517, +0.700] to +0.358 over a blind one on four open-weight backbones served with thinking disabled, and on the two Qwen backbones the blind inspector changes no item at all. That head-to-head is exploratory. One deterministic fault enters the first stage, and we re-expose the original problem to $k = 0,\dots,3$ of the downstream stages with agents, items, fault and topology held fixed, on 120 gsm_hard items per arm at temperature zero. Accuracy under fault rises on four of four backbones, from +0.233 to +0.392, the largest Holm-adjusted $p$ being $2.1\times10^{-6}$. A registered kill test rules out tokens. Blanking every word holds the word slots fixed, and retention tracks the visible fraction on four of four, climbing from 0.221 to 0.692 on the primary. Those two families are confirmatory and everything else here is exploratory. The interaction excludes zero on two of four backbones under the registered pipeline, four of four under a three-stage pipeline, and three of four under full message history, the primary at +0.317. On Llama-3.1-8B the fault carries no detectable cost at any dose, so the other three carry every claim about what a fault costs. Re-derivability also sets what the architecture costs, and no decomposition we measured reliably beats one direct call. With no fault injected the registered pipeline loses to that call by -0.267, -0.125 and -0.317, and on Phi-4 reads +0.058 at $p = 0.118$, which the test fails to separate from zero. The repair that works is cheap and front-loaded: the first re-grounded stage buys +0.394 of matched retention for +59.8 tokens per item on Qwen3-14B, and the stages after it buy nothing.