Using Grounded Theory for Agent Behavior Analysis at Scale

Understanding agent behavior requires methods that scale to thousands of trajectories and surface new patterns in long, often unfamiliar tasks where pre-built classifiers fall short. We propose to bring grounded theory into agent trajectory analysis: a six-decade-old qualitative method from the social sciences, with a principled saturation criterion and an auditable trail from data to theory. We propose AutoTraceGT (Automated Trace analysis through Grounded Theory), the first multi-agent pipeline that automates grounded theory on agent trajectories. It iteratively performs open, axial, and theoretical coding until saturation, producing a behavioral taxonomy tailored to each task. Across six trajectory corpora, AutoTraceGT produces codebooks that recover 73-91 percent of the failure modes in human-annotated taxonomies and surface additional patterns that those taxonomies miss. The emergent theoretical narrative aligns with prior expert accounts. Used as a deductive feature space, the codebook outperforms zero-shot and few-shot LLM baselines on downstream failure prediction. These results suggest Grounded Theory offers a scalable analytic tool for ML researchers and agent developers studying what agents actually do.

Last Step Matters: Early Uncertainty Cannot Predict Failure in Long-Horizon Agents

Early failure prediction is important for long-horizon agents, as it enables timely intervention and can reduce inference and tool-use costs. Uncertainty quantification, such as verbal confidence and perplexity, offers a promising approach to detecting agent failures; however, it has not been explored whether these signals retain their discriminative power during the intermediate stages of long-horizon execution. We evaluate mainstream uncertainty signals on deep-research tasks and find that verbal confidence reliably distinguishes failures at trajectory completion, achieving a mean AUROC of 0.85, whereas all evaluated signals offer limited predictive value earlier in execution, with none exceeding a mean AUROC of 0.60 at 50% trajectory progress. We identify an underlying mechanism explaining this gap: path switching, where agents frequently abandon their current search direction in-trajectory, breaking the link between early signal and final outcome. These findings challenge the assumption that intermediate uncertainty can reliably guide early intervention. They also motivate a practical recommendation for agent harnesses in deep-research settings: use final-step confidence to decide whether to restart, an approach that our experiments find more effective than in-trajectory intervention.

Automata from Agent Traces: Failure and Next-Step Prediction

LLM-based agents execute multi-step tasks, but their behavioral structure remains opaque: long unstructured traces resist the safety auditing and runtime monitoring that deployment requires. Existing approaches operate per-trace or success-only, so they miss the cross-run topology that links next-step and failure prediction. To recover that shared structure, we collapse an entire trace corpus into a single, compact finite-state machine (FSM) that serves as a structural substrate for the otherwise unpredictable behavior of LLM agents. Across twelve public datasets, the FSMs are compact (7-43 states), replay held-out data at >=0.997 fitness with near-identical topology across splits, and build in milliseconds. This substrate addresses both prediction goals. For next-step prediction, FSM-state context outperforms Agent Workflow Memory on every ground-truth-matched dataset. For failure prediction, per-state behavioral features reach held-out AUROC up to 0.94, and an online monitor ranks failing runs above passing ones from a partial trace, triggering early stopping well before completion. Behavioral topology thus appears shaped more by the deployment harness than by the LLM, providing a model-agnostic structural primitive for safety auditing and runtime monitoring.

$TCP_α$: Margin-Controlled Confidence estimation for reliable Music Information Retrieval

Deep neural networks are often overconfident, assigning high confidence even to incorrect predictions. Consequently, users lack a reliable signal for deciding when a prediction can be trusted. Post-hoc confidence estimation addresses this by training a lightweight auxiliary head over a frozen classifier. Existing targets, however, suffer from inherent ambiguity: they assign overlapping confidence values to correct and incorrect predictions, while errors near the decision boundary receive confidence scores indistinguishable from correct predictions. In this work, we propose $TCP_α$, a novel confidence target that resolves these limitations by introducing a margin-controlled penalty for misclassified samples. We prove that $TCP_α$ guarantees complete separation between the target values of correct and incorrect predictions, with a separation margin that is independent of the number of classes and increases monotonically with the penalty parameter. Since accurate classifiers naturally produce very few errors, learning these targets results in a severely imbalanced regression problem. We therefore present a systematic study of training strategies for learning under this imbalance and identify an effective training configuration through extensive ablation studies. We evaluate the proposed approach on rāga identification, investigate its robustness under domain shift, and further validate it on frame-wise ornamentation detection without modifying the selected configuration. Across all settings, $TCP_α$ consistently outperforms existing confidence targets for failure prediction. Rejecting only the least-confident 8\% of predictions improves the base model's macro-F1 from 0.89 to 0.98, while fine-tuning the confidence head with only 5\% labeled samples from a new corpus effectively restores performance under domain shift.

ContactGuard: Pre-Contact Execution Monitoring with Action-Conditioned Latent World Models

Contact-rich manipulation failures are often detected only after the robot has committed to contact. This is especially limiting in wrist-camera setups: close gripper--object views help observe contact, but a poor approach may already push, miss, slip, or disturb the object before conventional detectors react. We introduce \emph{ContactGuard}, a pre-contact execution monitor for chunked visuomotor policies. Given the policy's planned action chunk, ContactGuard predicts its short-horizon consequence in latent visual space and aborts if the predicted future latent indicates likely failure. Its latent world model is trained from unlabelled robot trajectories to predict compact multi-view visual embeddings under planned actions, avoiding pixel-level video prediction. A lightweight failure probe is then trained from a small labelled set of pre-contact clips. At deployment, ContactGuard anchors prediction before an imminent contact event, rolls the model forward under the policy's own actions, and verifies the predicted post-contact latent. Across real-world contact-rich manipulation tasks, ContactGuard predicts failure more accurately than direct and corrupted-action ablations, and transfers to live robot as a pre-contact abort signal without modifying the underlying policy.

Fail-Fast, Restart-Smart: Early Failure Prediction and Restart for SWE Agentic Tasks

Software engineering (SWE) agents resolve repository-level issues through long trajectories that grow increasingly expensive as context accumulates. Failed runs tend to be longer and exhibit redundant exploration or looping, suggesting that some failures may be detectable before completion. Early termination, however, risks interrupting trajectories that would otherwise succeed; conversely, an unsuccessful trajectory may still contain useful repository edits. We present FailFast-RestartSmart, a two-stage controller for a single active trajectory. FailFast is a lightweight 0.6B monitor trained with terminal and dense fail-to-pass supervision to predict failure from observable prefixes without policy logits or hidden states. Upon an alarm, RestartSmart launches a fresh same-policy rollout without prior prompt history and offers the interrupted repository diff as an optional overlay that the agent may inspect, apply, or discard. On SWE-bench Verified, a monitor trained solely on Qwen3.6-27B trajectories transfers to three other policies, including a closed-API model, and saves 14.6%-20.4% of execution tokens at a target 5% false-positive rate; on Qwen3.6-27B, its 20.4% saving exceeds the 12.5% achieved by our per-step AgentStop adaptation. At a target 25% false-positive rate, RestartSmart raises Qwen3.6-27B resolution from 66.6% to 71.8%, whereas cold restart reaches only 66.8%. Together, these results support early termination with sequential same-policy recovery.

Understanding Online Failure Prediction in Linux Through Complementary Multi-View Explainability

Accurate Online Failure Prediction (OFP) has been shown to be feasible in Operating Systems (OSs) settings, but prediction alone is not sufficient for practical adoption. Without diagnostic insight, operators have limited basis to trust alerts or decide how to respond. Moreover, even when predictive accuracy is high, it is often unclear whether models are capturing meaningful failure processes or merely exploiting workload-specific noise and incidental correlations in telemetry. This paper reports a practical experience building and evaluating an explainable OFP pipeline for Linux OSs. We combine consensus-based feature selection for detection with temporal onset analysis, subsystemlevel causal analysis, and complementary diagnostic mechanisms to support failure interpretation. Evaluated under strict crossworkload conditions with frozen training artifacts, it achieved 91-94% detection on unseen workloads without retraining, while maintaining false alarm rates below 1%. However, failure mode diagnosis proved substantially more sensitive to workload shift, and several diagnostics mechanisms showed limited effectiveness for specific failure types. Our experience highlights three main lessons: i) detection generalizes more robustly than diagnosis across workload changes; ii) early-warning capability depends strongly on the failure mode, ranging from 38 to 215 seconds in our study; and iii) unseen failure modes are not reliably diagnosable from related training modes alone, providing 0% accuracy under Leave-One-Mode-Out (LOMO) evaluation. Taken together, these results show the value of complementary explainability mechanisms for interpreting accurate failure predictions, revealing when predictive signals reflect transferable failure structure and when diagnostic generalization breaks down under workload variation.

Untangling Co-Drift: Proactive Multi-Intent Failure Prediction and Root-Cause Disambiguation for Self-Driving Networks

The vision of self-driving networks that monitor, reason, and act upon themselves with minimal human intervention relies on tightly coupled monitoring, analytics, and actuation functions. In this work, we treat these functions as three operational macro-intents: continuous telemetry, real-time analytics, and programmatic actuation, and formalize the health of each function as an intent that the network must continuously satisfy. A critical, yet underexplored, challenge stems from the causal coupling among these intents, where a singular fault within one macro-intent propagates as a co-drift and subsequently triggers cascading, symptomatic anomalies across the remaining intents. This ambiguity makes it exceedingly difficult for existing, reactive approaches to distinguish the true root-cause intent from symptomatic victim intents, and their reliance on threshold-crossing detection leaves insufficient time for proactive remediation. We introduce MILD, a novel framework that reformulates intent assurance from reactive drift detection to proactive failure prediction. Grounded in our three-macro-intent formulation of the self-driving control loop, MILD employs a teacher-augmented Mixture-of-Experts architecture with a hybrid objective that jointly optimizes intent failure prediction and root-cause attribution. MILD enables KPI-level diagnostics via SHAP explainability and dynamic intent failure urgency estimation via multi-horizon modeling. Our extensive evaluation of MILD across three environments of increasing realism, from a controlled statistical benchmark, to a microservices application, to an SDN-based edge-to-cloud testbed, demonstrates that MILD achieves high failure detection rates, strong remediation lead times, and accurate intent-level root-cause disambiguation. This positions MILD as a practical enabler of closed-loop assurance in next-generation autonomous networks.

The Knowing-Saying Gap: When Probes See Errors that Confidence Misses

Linear probes detect corrupted context in language models with near-perfect accuracy, yet this does not translate into reliable failure prediction. The result is a dissociation with direct implications for deployment monitoring. Across multi-hop arithmetic chains, probes that detect corruption turn out to be uninformative about final answer correctness; models forced into structured confidence formats collapse to two values with indistinguishable error rates; and probe persistence across hops fails to separate correct from incorrect outcomes, refuting our pre-registered "persistence beats peak" hypothesis. This pattern of knowing but not saying generalises across model families including reasoning models. As a real-time monitor, probe-based interventions are sharply model and error-type dependent: branch-and-pick is net-positive across models and uniquely non-breaking on Llama-3.1-8B (4 rescued, 0 broken), while reprompt and replace-prior break correct traces at roughly the rate they rescue wrong ones. Probe-based monitoring is a necessary complement to verbalised confidence, but no single intervention dominates, and the deployable answer is model-aware, error-type-aware routing.

PREFAIL: Identifying Precursors to Failures in Robotic Lift-and-Place Tasks to Improve Task Execution Performance

Non-prehensile manipulation enables flexible material handling with part carriers, but friction-based support makes high-speed motions failure-prone, while slower operation increases cycle time. Proactive failure prediction is therefore essential for efficient and reliable performance, yet existing approaches remain limited by key constraints, including sensitivity to dynamic actions and high dependence on known policy structures. Furthermore, existing methods and datasets lack a precise characterization of the latest intervention time, leaving it unclear whether a detected failure can still be prevented through timely intervention. In this paper, we investigate lift-and-place tasks for non-prehensile material handling manipulation and propose a more effective approach to predicting precursors to failures (PREFAIL) by analyzing the relative motion of target objects with respect to the carrier. We further introduce a dataset that precisely identifies the latest intervention time for risky manipulations, enabling rigorous evaluation of whether a failure prediction is actionable. We validate our approach on both simulation and real-world datasets. Our experimental results demonstrate that PREFAIL substantially improves both the accuracy and timeliness of responses to failure precursors.

Early Warning Signals for OpenVLA Failure under Visual Distribution Shift

Vision Language Action models combine perception, language grounding, and control in a single policy, but their failures are hard to diagnose once visual conditions shift. We test whether OpenVLA feedforward activations contain linearly decodable information about near term task failure in LIBERO manipulation rollouts. The policy is fixed throughout. We log internal activations during execution and fit lightweight monitors after the rollouts are collected. Occlusion is the main controlled stress test. It reduces OpenVLA success from $57\%$ to $17\%$ over $100$ episodes per condition. Under this shift, a logistic probe at layer 16 reaches AUROC $0.972$ and AUPRC $0.352$ for predicting failure within a $15$ step horizon. It outperforms both a mean difference direction and an action disagreement baseline. A sparse layer sweep finds uneven decodability across depth: layer 16 is strongest among the tested layers, layer 8 remains informative, and layer 10 is weaker. To check whether the monitor is just an occlusion detector, we also evaluate color shift and camera jitter without refitting. Color shift produces no failures in this setting, so it is a benign control rather than a failure benchmark. Camera jitter does induce failures, and the occlusion trained monitor remains above random. The result is deliberately limited: OpenVLA internal states contain failure relevant structure under controlled perceptual shift, but these experiments do not establish a causal mechanism, task held out generalization, or a deployable recovery system.

SSH-Net: A Deep Neural Network for Predicting Failure Time Distribution Functions under Competing Risks with Application to GPU Data

Competing risks are commonly observed in engineering fields and can bring challenges to time-to-event data modeling when the application scenarios are complicated. Recently, deep neural networks have received great attention for prediction with competing risks, due to their flexibility and high learning capability. However, the complexity of neural network structure brings extra difficulty in hyperparameter tuning based on different data inputs. Additionally, when an engineered system has complex physical structures with multiple hierarchical levels, treating all structural levels as a single group of inputs may fail to capture critical information. To address the issues, we propose a Structured Segmented Hazard Deep Neural Network (SSH-Net) for failure time prediction under cause-specific competing risks framework. Our approach associates neural network structure with data structures, and allows different covariate groups to impact the failure prediction through separate sub-networks. The neural network is constructed based on a cause-specific competing risks model. The SSH-Net outputs cause-specific hazard functions, and utilizes the penalized log-likelihood as the loss function. The prediction accuracy of SSH-Net is validated through simulation studies by evaluating the Brier score, the area under receiver operating characteristic curves (AUC), and the root mean square error (RMSE) of the predicted cause-specific cumulative incident function. We further demonstrate the model's ability to predict failure time distribution functions using the Titan GPU failure time data.

Tri-Info: Generalizable, Interpretable Failure Prediction for VLA Models via Information Theory

Vision-Language-Action (VLA) models are increasingly deployed across diverse tasks, yet they remain black boxes whose physical interactions can cause irreversible harm, making generalizable and interpretable failure detection essential. We observe that successful and failed rollouts carry systematically different information-theoretic signatures. Building on this, we formalize VLA control as a closed-loop information pipeline and derive the Triple Information-theoretic (Tri-Info) signals that capture whether actions remain diverse, temporally consistent, and coupled to state transitions. Across six VLA models and three benchmark environments, Tri-Info matches the strongest baselines in-domain. Moreover, Tri-Info transfers across architectures, environments, and the sim-to-real gap without retraining, reaching 83\% accuracy on real-world tasks where prior detectors collapse to chance. This establishes Tri-Info as a simple yet powerful method that not only detects failures with strong cross-domain generalization, but also delivers interpretable diagnostics of the underlying failure modes.

Compositional Reasoning Depth Predicts Clinical AI Failure: Empirical Evidence Consistent with Transformer Compositionality Limits in Electronic Health Record Question Answering

Aggregate accuracy benchmarks conceal a systematic structure in how large language models fail at electronic health record (EHR) question answering: questions requiring more inferential steps produce disproportionately more errors. Motivated by theoretical results on transformer compositionality limits, we introduce a pre-specified hop-count taxonomy -- the number of distinct reasoning steps required to answer a clinical question from an EHR -- as a principled predictor of model failure. We annotate 313 clinician-generated MedAlign EHR question-answer pairs across four hop levels and evaluate 301 questions in a within-model ablation (claude-sonnet-4-6, zero-shot vs. extended thinking) and cross-architecture replications (gpt-4o and gpt-5.4-2026-03-05, zero-shot). All three models, spanning two providers and two OpenAI generations (GPT-4 and GPT-5), show monotone accuracy decline with hop count: Claude Sonnet zero-shot falls from 30.6% (hop=1) to 17.6% (hop=4) (Cochran-Armitage z=-2.30, p=0.011; OR per hop 0.72, 95% CI [0.56,0.92], p=0.008); GPT-4o replicates this (37.8% to 14.7%; OR 0.58 [0.45,0.75], p<0.001); and gpt-5.4-2026-03-05 confirms it (37.8% to 23.5%; OR 0.80 [0.66,0.98], p=0.027). A pre-specified context-sufficiency audit shows higher-hop questions are not differentially disadvantaged by EHR truncation (answerability 93-95% at hops 2-4 vs. 79% at hop=1), so the decline reflects compositional reasoning difficulty. Extended thinking did not significantly flatten the accuracy-depth curve across three reasoning conditions, and thinking-token usage scaled with hop count (r=0.31, p<0.0001), consistent with the predicted O(k) computational requirement. Hop count is thus a theory-motivated, cross-architecture predictor of large-language-model error on EHR question answering, with direct implications for deployment risk stratification of clinical AI.

Adversarial Concept Search: Predicting Compositional Errors From Feature Geometry

Humans cannot always intuit what scenarios are most challenging to LLMs. Hoping to capture challenging edge cases, developers either design problems to be difficult for humans or curate extensive benchmarks. What if we could instead anticipate which scenarios a model will fail on? In this paper, we use an LLM's representational geometry to predict which concept combinations it will fail on. We attribute this compositional failure to interference between salient features. In tasks that require systematic composition - toy programmatic settings, multihop reasoning, multilingual factual recall - we find that when a pair of concepts is encoded near-orthogonally, the model reliably composes them. When their linear encodings are close, producing interference, the model fails to compose them. Our method reliably anticipates failure modes across different compositional tasks, without evaluating specific inputs. These results lay the groundwork to use representational geometry to identify high-risk examples, construct targeted stress tests, and provide a scalable foundation for active learning in real-world deployment.