When Evidence is Sparse: Weakly Supervised Early Failure Alerting in Dialogs and LLM-Agent Trajectories

arXiv cs.CL Papers
weakly-supervised-learning early-failure-alerting dialog-systems llm-agents sparse-evidence attention-based-predictor alpha-stop

Summary

This paper proposes a two-stage approach for early failure alerting in dialogs and LLM-agent trajectories, addressing the challenge of sparse evidence by learning turn-level failure evidence from trajectory labels and using an attention-based predictor with a preference-conditioned stopping policy (α-STOP) to achieve controllable accuracy-earliness trade-offs.

arXiv:2606.05414v1 Announce Type: new Abstract: Early failure alerting requires deciding, while a dialog or agent trajectory is still unfolding, whether to flag it as likely to fail. This is challenging because supervision is typically available only as a trajectory-level success/failure label while alerts must be raised from partial interactions. Prior early-classification methods often bridge this gap by assigning the terminal label to every prefix, treating every turn as failure evidence. We hypothesize that this prefix-label assumption is poorly matched to multi-turn language interactions, where evidence of eventual failure is sparse and often delayed. In this paper, we introduce a two-stage approach that learns from this sparse evidence structure and uses the resulting risk estimates for controllable early alerting. Specifically, our attention-based failure predictor learns sparse turn-level failure evidence from trajectory labels and uses it to estimate failure risk from partial histories. We then pair this predictor with $\alpha$-STOP, a single preference-conditioned stopping policy that selects an accuracy-earliness operating point at inference time rather than training a separate trigger for each preference. Across five benchmarks spanning customer support, task-oriented dialog, persuasion, tool use, and planning, we first show that high-relevance failure evidence occupies only 4.7-11.3% of turns and first appears after 59.0-83.6\% of trajectories on average. We further show that the attention-based predictor improves Pareto-frontier quality (hypervolume) by 1-10\% over naive prefix supervision, and that the full system improves frontier quality by 3-42\% over state-of-the-art trigger policies while reducing training cost per operating point by 1-3 orders of magnitude.
Original Article
View Cached Full Text

Cached at: 06/05/26, 08:06 AM 

# When Evidence is Sparse: Weakly Supervised Early Failure Alerting in Dialogs and LLM-Agent Trajectories
Source: [https://arxiv.org/abs/2606.05414](https://arxiv.org/abs/2606.05414)
[View PDF](https://arxiv.org/pdf/2606.05414)

> Abstract:Early failure alerting requires deciding, while a dialog or agent trajectory is still unfolding, whether to flag it as likely to fail\. This is challenging because supervision is typically available only as a trajectory\-level success/failure label while alerts must be raised from partial interactions\. Prior early\-classification methods often bridge this gap by assigning the terminal label to every prefix, treating every turn as failure evidence\. We hypothesize that this prefix\-label assumption is poorly matched to multi\-turn language interactions, where evidence of eventual failure is sparse and often delayed\. In this paper, we introduce a two\-stage approach that learns from this sparse evidence structure and uses the resulting risk estimates for controllable early alerting\. Specifically, our attention\-based failure predictor learns sparse turn\-level failure evidence from trajectory labels and uses it to estimate failure risk from partial histories\. We then pair this predictor with $\\alpha$\-STOP, a single preference\-conditioned stopping policy that selects an accuracy\-earliness operating point at inference time rather than training a separate trigger for each preference\. Across five benchmarks spanning customer support, task\-oriented dialog, persuasion, tool use, and planning, we first show that high\-relevance failure evidence occupies only 4\.7\-11\.3% of turns and first appears after 59\.0\-83\.6\\% of trajectories on average\. We further show that the attention\-based predictor improves Pareto\-frontier quality \(hypervolume\) by 1\-10\\% over naive prefix supervision, and that the full system improves frontier quality by 3\-42\\% over state\-of\-the\-art trigger policies while reducing training cost per operating point by 1\-3 orders of magnitude\.

## Submission history

From: Avinash Baidya \[[view email](https://arxiv.org/show-email/c3acd6bd/2606.05414)\] **\[v1\]**Wed, 3 Jun 2026 20:28:27 UTC \(2,655 KB\)

Similar Articles

From Confident Closing to Silent Failure: Characterizing False Success in LLM Agents

arXiv cs.LG

This paper characterizes 'false success' in LLM agents, where agents claim task completion despite environment state showing otherwise, finding it accounts for 45-75% of failures across benchmarks. LLM judges fail to detect this reliably, while lightweight TF-IDF detectors achieve high AUROC with much lower latency, suggesting production monitoring should use calibrated detectors instead of LLM judges.

AgentForesight: Online Auditing for Early Failure Prediction in Multi-Agent Systems

arXiv cs.CL

This paper introduces AgentForesight, a framework for online auditing and early failure prediction in LLM-based multi-agent systems. It presents a new dataset, AFTraj-22K, and a specialized model, AgentForesight-7B, which outperforms leading proprietary models in detecting decisive errors during trajectory execution.