Overview

Test-time adaptation studies how pretrained models can adjust to changing data during deployment, improving reliability when target labels or retraining are unavailable. Common approaches use entropy minimization, pseudo-labels, normalization or feature-statistics alignment, and lightweight prompts, adapters, caches, or memory; gradient-free methods aim to reduce compute and memory costs. Work spans image corruptions and domain shifts, medical imaging, time-series forecasting, language-model reasoning, and robotics. A central tension is that noisy self-generated signals can cause error accumulation or forgetting, so methods must determine when adaptation is useful and how to evaluate its benefits beyond average accuracy.

Latest papers

342 papers40 papers in the last four weeks, down 17% on the four weeks before. 0.3% of all new papers.

Nov 13, 2026

What Balanced Accuracy Cannot Identify in Neural-Signal Decoding under Test-Time Adaptation

Adaptation for neural-signal classification is usually transductive: each batch is scored after updating on it, and the gain depends on the adapted model [4]. This paper asks what an aggregate change fails to characterise inside the cohort. Full test-time adaptation (full_ttt) was evaluated in 11 dataset-encoder cells (four each on Siena/EEG and PTB-XL/ECG, three on Sleep-EDF/EEG), with splits and source-model evaluation fixed; the subject was the unit, exact sign tests were Holm-corrected and conditional on the shared stream, and bootstrap intervals accompanied a dependence audit. Two results of different kinds follow. In the binary case balanced accuracy gives a one-class decoder the same value as one reporting both classes at 50%, so the aggregate cannot identify a class that stopped being reported; a class covering a tenth of a subject's samples went unpredicted in 26-47 of 112 subject-cells, against 7-9 for the controls, which is a descriptive count. With cohort, split and update rule fixed, the within-subject class contrast follows the encoder, tested directly rather than inferred from tests against zero: two of six Siena encoder pairs survive the dependence readings at their nominal reference and one at a fold-level reference, while two of ten Sleep-EDF stage pairs pass the run-clustered gate at its nominal reference; aggregate and contrast named different encoders in 10 of 11. The protocol bounds both results, since a stream-interleaving effect reverses sign with state sharing (+3.5 to -2.4 pp). An aggregate score can hide individual harm, as [1] showed, but it cannot distinguish a uniform loss from a deleted class; class-level change should therefore be reported alongside the aggregate and subject-level change alongside it, with both results conditional on the cell, the rule and the stream construction.
Nov 12, 2026

DEDCA: Test-Time Adaptation for Generalized AI-Generated Image Detection

AI-generated image detectors face significant challenges when deployed in real-world environments, particularly when test samples are produced by unseen generators that deviate from the source training distribution. Existing detectors often rely on a single type of evidence, such as semantic representations or low-level generation traces, making them vulnerable when the corresponding cue becomes unreliable. Moreover, adapting detectors to newly emerging generators usually requires target labels or retraining, which is costly and impractical. To address these challenges, we propose Dual-Evidence Disagreement-Constrained Adaptation (DEDCA), a test-time adaptation framework for generalized AI-generated image detection. DEDCA constructs a dual-evidence detector by combining CLIP-based semantic evidence with trace evidence extracted from shuffled entropy and residual statistics, and introduces a reliability-aware gate to perform sample-specific adaptive fusion. Our key idea is to exploit the disagreement between the two evidence streams as an unlabeled adaptation signal, enabling the detector to adjust to target-domain shifts without relying on target labels or blindly trusting a single prediction stream. Experiments show that DEDCA achieves superior performance on the GenImage benchmark by outperforming state-of-the-art AI-generated image detectors and test-time adaptation methods with 92.0% average accuracy.
Nov 11, 2026

Adapting Monocular Depth Estimation in Novel Scenes through Disparity Refinement and Robot Ego-Motion

The geometric understanding of a scene is a crucial component for safe and efficient robotic operations in novel and unstructured environments. The recent emergence of foundation models for Monocular Depth Estimation (MDE) has provided an opportunity to enrich robotic perception capabilities with dense long-range depth predictions, replacing expensive onboard hardware (e.g., LiDARs and stereo cameras). However, these offline models are primarily trained on affine-invariant depth representations, and their generalization to novel scenes remains unclear. In this work we present EM-ViT, a monocular depth estimation model that learns to adapt online relative depth inputs from a foundation model prior to the metric scale with observations collected during robotic operations. Motivated by the stereo camera parameterization, our method decomposes depth prediction via dense disparity refinement to recover the scene's relative structure and an effective baseline scalar prediction to determine the missing metric scale. The learning process leverages sparse LiDAR points and, optionally, the robot's Ego-Motion (EM) to rapidly train a lightweight Vision Transformer (ViT), while an internal gradient matching loss preserves the fine-grained semantic structure of the input disparity in real-time. We demonstrate the efficacy of our method on publicly available sequences, such as KITTI, and show that our online learning pipeline can adapt to novel scenes faster and with higher accuracy than existing MDE foundation models. The code and supplementary materials will be publicly available at https://emvit-mde.github.io.
Nov 9, 2026

Library of Hands: Fix the Hands, Not the World Model

A world model predicts how a scene changes after an action, and a planner uses it to search for actions that reach a goal. If the camera moves or the controller changes, the same command moves the robot differently: when its commands are turned by 90 degrees, a frozen planner with the LeWM world model that succeeds in 77.5% of episodes succeeds in only 7.8%. Current fixes retrain the model, need training data, or forget a set-up when it returns. We propose Library of Hands (LoH), a training-free method that keeps the world model frozen and corrects only the map from commands to applied actions, which we call the hand. LoH reads the hand through the frozen model, with no extra sensor: it finds the action that best explains the change between two camera images (a reading). It stores every hand it identifies, so a returning set-up is recognised rather than re-learnt. In closed-loop tests, LoH more than doubles success on three simulated tasks, from 28.6% to 62.3%, and loses only 0.6 percentage points when nothing changes. It also works, unchanged, with two other world models and on two of three models pretrained on real robot data, and its gain grows with the quality of its readings.
Nov 8, 2026

Actions Should EARN Their Credit: Test-Time Adaptation and Route Consolidation for Frozen LLM Agents

LLM agents can use feedback from repeated attempts to improve their decisions without updating model parameters. However, positive action credits can continue to favor actions that repeatedly achieve the same score, even when higher scores are achievable. Under a fixed interaction budget, trying new actions in the final attempts can also lower performance. We introduce EARN, a method that uses past experience to guide further progress and retain earlier performance through conditional reuse of effective action sequences. EARN reduces action credits for reaching scores repeatedly attained in earlier episodes and assigns larger credits to action segments that reach higher scores. These credits adjust the probabilities of LLM-proposed actions. During the final episodes, EARN directly replays stored actions when the current context matches the recorded context and the actions remain available, bypassing LLM candidate generation. We evaluate performance at the end of the interaction budget using the average score over the final four episodes. Across Jericho, ScienceWorld, and Crafter, EARN achieves the best score on this metric in 31 of 33 task--budget settings among the evaluated methods, including ties. Across the evaluated Jericho budgets, replay bypasses 78.9--86.3% of candidate-generation calls in the final four episodes, averaged across games.
Nov 8, 2026

Let the Gallery Learn Itself: Training-Free Iterative Side-View Face Recognition

What if a face recognition system could learn from every encounter without a single gradient update? We present Gallery Self-Evolution (GSE), a training-free framework where the gallery itself adapts at test time. The core mechanism---dedup-FIFO curation---archives novel appearance samples after successful recognition while permanently locking the enrollment template, bounded at 30 embeddings per identity. On CFP-FP and IJB-C benchmarks, GSE closes the side-view gap by +23 points at 45° yaw and boosts TAR@FAR=0.001 by +9.4 points, all at CPU-level cost. No retraining. No GPU. Just a gallery that grows smarter every time it meets you.
Nov 6, 2026

Toward the Era of Agentic Embodied Intelligence

Foundation models increasingly guide robot actions, yet sustained autonomy requires efficient control and failure recovery. We introduce A2-MEAS, an embodied multi-agent framework built around code as a scaling interface. Learned correction codes combine corrective language, spatial evidence, execution history, and action proposals. MetaBrain provides training supervision and runtime guidance. A correction encoder and adapter absorb this knowledge into policy weights during training, while fresh codes adapt behavior with frozen weights at deployment. The action model, A2-WLA, adds motion-aware history and a latent world model to a vision-language-action policy. Surprisingly, this simple design with action pretraining and task specialization reaches 0.5755 Q-score on the 50-task BEHAVIOR 2026 public subset before online correction, 2.21x the 2025 winner's reported 0.2605 across different benchmark editions and training configurations. On all 100 BEHAVIOR 2026 public tasks, A2-MEAS achieves 0.6411 Q-score and 45.40% task success. On RoboTwin 2.0, A2-WLA achieves 93.98% average success, rising to 95.05% with the complete system. Across 12 real-robot tasks in five household settings, A2-WLA achieves 57.92% average success versus 26.67% for pi0.5. In long-horizon pour-over coffee preparation, a history-conditioned adapter raises post-failure completion at two filter-handling steps from 14% to 94% and from 8% to 82%, compared with A2-WLA retries. Our training infrastructure delivers 427.1 samples per second, 4.03x the throughput of a pi0.5 reference implementation on matched hardware. We will release the system, models, training scripts, and infrastructure to support efficient embodied learning.
Nov 4, 2026

ReTailor: A Residual Feature Tailoring Framework for Real-Time Cross-Sensor Pansharpening Adaptation

Recent pansharpening models achieve strong fusion performance on trained sensors, yet often generalize poorly to unseen sensors. Existing cross-sensor adaptation is limited by target-sensor retraining, which requires target-sensor training pairs and a dedicated training stage, and by per-image zero-shot adaptation, which is computationally expensive due to whole-network optimization for each test image. To address these problems, we propose ReTailor, a real-time sequential cross-sensor adaptation framework. Only a lightweight residual Feature Tailor is rapidly optimized under self-supervised, physics-aware constraints to correct cross-sensor shifts without modifying the well-learned fusion prior. For efficient and effective deployment, ReTailor combines content-aware patch selection, which identifies a few informative patches per image for optimization at negligible cost to improve both quality and efficiency, with rolling residual bootstrap, which inherits sensor-related knowledge across the image stream to further improve quality. Extensive experiments on cross-sensor adaptation settings show that ReTailor mitigates cross-sensor degradation both effectively and efficiently. Moreover, we compare ReTailor with Low-Rank Adaptation (LoRA), a widely used method for fine-tuning, to demonstrate the advantages of our approach. We further evaluate the core principle of ReTailor on two additional remote sensing tasks: super-resolution pansharpening and image segmentation, where it consistently improves performance. In general, ReTailor provides a highly effective and efficient strategy for deploying pansharpening models under real-world cross-sensor degradation, while also offering useful insights into managing cross-sensor degradation in other remote sensing tasks. Code will be released at https://github.com/JingZe-Liang/ReTailor .
Nov 3, 2026

HARP: Training-Free Dual-Profile Agentic Communication for LLM-Based Recommendation

In LLM-based recommendation, memory acts as a user-level agent in a memory-augmented backbone. Yet users and items remain asymmetric: only the user is given a structured profile; it enters a bandwidth-limited reranker prompt; and it is frozen after warmup. We recast LLM-based recommendation as an agentic-communication process in four stages: BUILD a self-description, TRANSMIT it along the collaborative graph, UNDERSTAND the other side, and SYNC it from the next interaction. This formulation reveals three unresolved communication challenges: the item side never produces a comparable self-description (one-sided BUILD); profile-level signals enter only the bandwidth-limited reranker prompt rather than retrieval, where they would not compete for context tokens (prompt-trapped UNDERSTAND); and the user's description is frozen after warmup with no principled update rule (absent SYNC). We instantiate the missing stages as HARP (Homophily-aware Agentic Recommendation via dual Profiles), a training-free, three-operator addition to a frozen LLM backbone: a rule-extracted symmetric item profile (zero extra LLM calls), a channel-decomposable profile-level homophily H(phi_u, phi_i) consumed at retrieval time rather than inside the reranker prompt, and a single-pass reflective update realising in-context profile evolution without any gradient step. On four public benchmarks (Yelp, Books, MovieTV, Goodreads), HARP lifts Hit@1 over the strongest published baseline by 6.9%-24.0%, with family-wise Holm-corrected p < 1e-9 on every dataset. Ablations and a per-case attribution study show that the operators contribute complementary gains, with the retrieval-side gate as the main source; we also report the trade-off introduced by the reflective pass. We formulate agentic communication and provide a training-free instantiation. Code is available at https://github.com/Jove-Zhen/HARP.
Nov 3, 2026

Uncertainty-Aware Prototype Learning for Open-Set Test-Time Adaptation

Open-set test-time adaptation (TTA) aims to adapt source-trained models to unlabeled test streams where covariate-shifted in-distribution (csID) samples and covariate-shifted out-of-distribution (csOOD) samples may coexist. Existing TTA methods are typically designed under a closed-set assumption and therefore lack explicit mechanisms for reliably distinguishing csID from csOOD samples. Consequently, csOOD samples may be mistakenly included in confidence-enhancing adaptation, producing harmful gradients that degrade known-class performance. To address this, we propose Uncertainty-Aware Prototype Learning for Open-Set Test-Time Adaptation (UAPA), a plug-and-play method for reliable open-set adaptation. Specifically, we first introduce Uncertainty-Driven ID/OOD Separation, which combines a classifier-prototype score with a mini-batch-normalized evidence-derived uncertainty score to distinguish csID from csOOD samples. We then propose a Dynamic Prototype Refinement strategy that periodically refines class-specific target prototypes using buffered csID features to progressively follow the estimated class directions. Lastly, we introduce Contrastive Prototype Entropy Adaptation, which applies entropy minimization to csID samples and entropy maximization to csOOD samples, while prototype-level contrastive learning aligns csID features with their corresponding target prototypes and reduces the maximum similarity between csOOD features and known-class prototypes. Empirically, UAPA outperforms the strongest baselines on the CIFAR benchmarks and Tiny-ImageNet-C, improving average OSCR by 2.08 and 1.28 percentage points, respectively.
Nov 3, 2026

Bridging CLIP with DINO: Cross-Modal Information Maximization for Online Test-Time Adaptation

Vision-language models such as CLIP enable zero-shot classification from arbitrary class names, but under test-time corruptions, their visual encoder projects distorted features onto incorrect textual anchors with high confidence: a phenomenon we term modality bias. Standard entropy minimization reinforces these overconfident errors, which creates a confirmation-bias loop and collapses the predicted label distribution. We propose an online test-time adaptation (TTA) method that breaks this loop at three levels. At the representation level, a frozen self-supervised backbone (e.g., DINO v2) is combined with CLIP through a cross-covariance logit ensemble. CLIP text prototypes are projected into the feature space of this text-free backbone, so the self-supervised branch classifies native image features and is less affected by the text bias of CLIP. At the optimization level, a gated test-time information-maximization objective pairs per-sample sharpening with batch-level diversity, while an adaptive gate excludes unreliable samples from the gradient. At the prediction level, an online logit adjustment corrects the residual class-distribution skew. CDIM outperforms the compared CLIP-based online TTA baselines on ImageNet-C, CIFAR10-C, and CIFAR100-C, and remains competitive in continual and mixed-domain TTA. Extensive analyses and ablation studies verify the contribution of each of its components. Our code is available at https://github.com/B1ueC1oud/CDIM
Nov 2, 2026

Calibrating Multimodal Test-Time Adaptation under Reliability and Availability Modality Shifts

Multimodal test-time adaptation (TTA) enables pretrained multimodal models to accommodate distribution shifts during inference using only unlabeled test data, without recourse to the original training data. Compared with unimodal adaptation, multimodal adaptation is further complicated by the coexistence and interaction of multiple representation perspectives. This complexity is particularly pronounced under modality reliability shifts, which preserve modality availability but degrade information quality, and modality availability shifts, which render the corresponding modality entirely inaccessible. Category-conditional Gaussian modeling offers a natural starting point for handling such distributional variations, having proven effective in unimodal test-time adaptation for characterizing representation distributions, yet its extension to multimodal settings requires a more elaborate formulation. To this end, we propose Adaptive Probabilistic Gaussian Calibration (AdaPGC), which extends category-conditional Gaussian modeling across modality-specific and multimodal fusion representations for probabilistic calibration under multimodal distribution shifts. AdaPGC continuously updates category-conditional Gaussian models over modality-specific and fusion representations, using the resulting probabilistic predictions to calibrate source model predictions and guide reliability based rectification, while exploiting their joint distributions to recover unavailable fusion representations. Extensive experiments under diverse modality corruption and missingness conditions demonstrate that AdaPGC effectively improves robustness under both modality reliability and availability shifts. Our source code will be released at https://github.com/XuJinglinn/AdaPGC-plus.
Nov 1, 2026

MIRA: Mutual Information guided calibration for Reliable Test-Time Adaptation

Test-time adaptation (TTA) updates a deployed model using only unlabeled target data, but standard entropy-based methods often rely on deterministic confidence and can reinforce overconfident errors under distribution shift. We propose MIRA, Mutual Information guided calibration for Reliable Test-Time Adaptation, a source-free online adaptation method that uses lightweight last-layer Gaussian perturbations to probe prediction reliability. For each target batch, MIRA reuses a single feature extraction pass and samples perturbed classifiers around the source classifier, forming an efficient local predictive distribution. Rather than fitting a Bayesian posterior, MIRA treats the perturbation scale as a local probing radius and selects a batch-specific scale using mutual information (MI) and prediction agreement. The resulting perturbation-sensitivity signal identifies predictions that are confident and locally stable under classifier perturbations, while suppressing brittle predictions with high stochastic disagreement. MIRA then performs adaptation with an MI-confidence weighted entropy loss, which down-weights perturbation-sensitive or low-confidence samples before they drive entropy minimization. This weighted loss reduces the influence of brittle predictions that can otherwise cause harmful adaptation under distribution shift. The method requires no source data, retraining, model ensembles, or offline posterior fitting, and can be applied as a plug-and-play modification to entropy-minimization TTA. Experiments on distribution-shift benchmarks show improvements in accuracy, calibration, and negative log-likelihood, suggesting that mutual-information-guided perturbation probing and weighted adaptation yield more robust online TTA than deterministic confidence.
Nov 1, 2026

Leveraging 2D Foundation Priors for Generalized Category Discovery under Test-Time Adaptation in 3D LiDAR Segmentation

In autonomous driving, 3D LiDAR semantic segmentation is expected to adapt to evolving environments and recognize previously unseen categories, making generalized category discovery under test-time adaptation (GCD-TTA) an important problem. However, existing works treat GCD and TTA separately, while the limited semantic information in 3D point clouds makes GCD-TTA under domain shift more prone to feature collapse when combining GCD with TTA. To address this issue, we propose a two-stage framework for GCD-TTA in 3D LiDAR semantic segmentation. Specifically, in the first stage, we adaptively aggregate local 3D features according to geometric density variations and obtain reliable pseudo labels under a global class-wise prototype consistency constraint. In the second stage, we cast category discovery as an optimal transport assignment process for globally consistent grouping of novel categories. Meanwhile, semantic features from 2D vision foundation models are fused with 3D point cloud features to enrich 3D semantic information and promote a more separable feature space. Extensive experiments on Synth4D to SemanticKITTI, SynLiDAR to SemanticKITTI, and Synth4D to nuScenes demonstrate the effectiveness of our framework. Under a unified evaluation protocol, our method achieves the best overall mIoU on 9 out of 10 benchmark splits and improves the overall mIoU by up to 12.21 points over the strongest re-implemented baseline.
Oct 29, 2026

Agent-Wise Test-Time Normalization for Camera-Based Cooperative Perception with Corrupted Collaborators

Intermediate-fusion cooperative perception lets a vehicle detect objects beyond its own field of view by fusing features shared by nearby agents. When only a collaborator's cameras are degraded by fog, motion blur or exposure errors, the features it shares drift away from the training distribution while the ego input stays clean, and cooperation can become worse than not cooperating at all. Test-time batch-normalization adaptation, the standard training-free remedy, estimates statistics over the pooled batch of all agents and therefore normalizes both the clean ego and the corrupted collaborator with a mixture of two distributions. We show that this mixing harms even the clean ego: when the ego shares BN statistics with a corrupted collaborator, its detection from its own features drops by 3.58 to 6.68 [email protected] points although its cameras are untouched. We propose AgentNorm, an agent-wise test-time normalization that i) estimates the normalization statistics of every collaborator from its own current observation and ii) anchors the ego to its source statistics, so that the ego path remains exactly the original model. AgentNorm has no learnable parameters, needs no gradients or labels, is independent of the order of test frames, and runs locally on each agent without changing the transmitted features. On camera-based OPV2V with four fusion checkpoints and collaborator-only corruptions, AgentNorm raises [email protected] under severity-5 fog from 29.42-37.25 to 43.32-58.86 and outperforms pooled normalization by 5.53-12.94 points under fog and motion blur, at an extra latency of 2.3 ms per frame. On the real-world DAIR-V2X-C dataset, it outperforms pooled normalization in all ten infrastructure-corruption settings. We also report the cases it does not fix: occlusion, which removes image content rather than shifting feature statistics, and the case in which all agents share the same shift.
Oct 29, 2026

Timbre: Reliability-aware harmonic consensus for label-free multimodal test-time rexinforcement learning

Label-free test-time reinforcement learning adapts vision--language models using rewards derived from their own predictions. However, misinterpreted visual evidence and sensitivity to textual formulation can produce a consistent but incorrect consensus that majority voting reinforces. Cross-view aggregation can expose such errors, but image and text transformations may themselves degrade task-relevant evidence to different extents. We introduce \method, a reliability-aware harmonic consensus framework for multimodal test-time reinforcement learning. \method constructs a 2×22\times2 grid combining original and perturbed images with direct-solving and reframing modes, selects pseudo-labels through weighted harmonic aggregation, and adapts modality-level weights using batch-averaged, temporally smoothed shifts in pseudo-label confidence as a heuristic reliability signal. We formulate consensus as a weighted χ2χ^2-barycenter problem whose unique minimizer is the normalized weighted harmonic aggregate, without requiring correct-answer invariance. An exact support--dispersion decomposition characterizes when harmonic aggregation preserves or corrects arithmetic rankings, while a sensitivity analysis establishes how modality reweighting changes pairwise ranking margins, informing interpretation of the heuristic updates. Across six benchmarks and three Qwen3-VL model scales, \method outperforms TTRV in 16 of 18 benchmark--model settings on adaptation instances and 17 of 18 settings on held-out instances under matched estimation and optimization rollout counts. These results demonstrate improved target-task adaptation and transfer to unseen instances without further updates.
Oct 28, 2026

Calibrating Track-Query Updates at Test Time for Multi-Object Tracking

Multi-Object Tracking (MOT) requires localizing multiple objects and maintaining their identities over time. Recent query-propagation trackers carry target-specific representations across frames and update them with current-frame image features, yet identity association remains difficult when visually similar targets overlap or cross. During such interactions, current-frame query updates may contain components directed toward nearby active tracks, which may affect identity association in later frames. To investigate this issue, we analyze track-query updates preceding identity switches (IDSWs), where the same target is assigned a different track ID. Our analysis shows that (i) propagated track-query representations change more abruptly before IDSWs, (ii) current-frame updates have larger positive projections along directions toward other active tracks, and (iii) bounding-box intersection over union (IoU) retrieves the counterpart track involved in a subsequent IDSW with high recall when that track remains active before the switch. Motivated by these findings, we propose Test-time Track-Query Calibration (TQC), a training-free and gradient-free method for query-propagation trackers. TQC calibrates the tracker output before propagation using a bounded correction estimated from update components directed toward nearby active tracks. It requires only track-query representations and bounding boxes available during inference, without retraining or architectural modification. Across nine DanceTrack and SportsMOT test settings, TQC improves HOTA and AssA in all settings and IDF1 in eight, with an average HOTA gain of +0.38 and limited inference overhead. Controlled variants on six validation settings further show that TQC's correction direction yields more consistent HOTA improvements than alternative directions with matched correction norms.
Oct 27, 2026

When Should We Intervene at Test Time?

Test-time adaptation helps a deployed model respond to changing inputs, but the same intervention can rescue one batch and damage the next. Updating normalization statistics, for example, can improve corrupted images yet fail when a batch contains mostly one class. Average accuracy hides these failures and cannot tell a model when to use an intervention. Target labels can guide this decision, but they are scarce and some choices are already clear. We introduce PORTAL, a controller that chooses when to deploy an intervention, which inputs to label, and when to stop querying. Each available prediction rule is a candidate, and its predictions are fixed before current feedback arrives. Every acquired label therefore evaluates all candidates on the same input. PORTAL combines these shared comparisons and directs queries toward candidates that can still overtake the current leader. It stops when the remaining labels cannot change the choice or when the possible accuracy gain no longer justifies another query. Our analysis links the required feedback to accuracy gaps and shared mistakes, explaining why some choices need much less evidence. Experiments show that this control turns an intervention that loses accuracy overall into a useful choice on selected batches. It prevents severe failures caused by changes in arrival order and preserves gains across complete corruption streams. When a fixed rule already works well, PORTAL uses little feedback to retain its performance. Periodic label checks reveal when past evidence needs refreshing and help the controller recover as conditions change. The acquired labels also train stronger candidates for future batches, and selection brings nearly all of their accuracy gains into deployment. The same feedback therefore improves both the choice made now and the predictions available to later batches.
Oct 27, 2026

Tail Cues, Principal Corrections: Plug-and-Play Rectification for Open-Set Test-Time Adaptation

Open-Set Test-Time Adaptation (OSTTA) operates on non-stationary target streams where covariate and semantic shifts coexist. Existing TTA methods often obtain adaptation signals by updating model parameters or maintaining mutable memories, which can be costly and vulnerable to unknown samples. We present TALOR, a lightweight, backward-free rectification module that uses the source linear head as a geometric anchor. The head-induced basis decomposes each normalized feature into high-gain principal coordinates and low-gain tail coordinates. TALOR estimates a soft-routed affine correction in the principal subspace, combining a regime-level principal bias with a tail-conditioned slope that captures sample-level tail-to-principal coupling. Using weighted ridge regression, it subtracts this drift estimate from the principal coordinates, reconstructs the feature and feeds it into the source head. On the ImageNet-C benchmark with six csOOD datasets, TALOR outperforms the next-best UniEnt by 2.2% H-score while running 2.5x faster and using only 12% of its GPU memory; as a post-correction plug-in, it further boosts COME to 68.4% H-score with only 2% additional memory overhead.
Oct 27, 2026

OmniGlimpse: Asymmetric Identity Memory for Online Multi-View Tracking

Multi-view tracking extends visual coverage beyond a single camera, providing complementary observations when targets are occluded or leave the view. This raises a question: how can observations from different views be matched to the correct identities without offline scene-specific training or camera calibration? Progress toward this goal remains limited because many offline formulations assume such deployment-specific knowledge, while online adaptation can amplify association errors. We study this problem as online multi-view identity maintenance (OMIM): targets are initialized once per view, with no further identity labels provided after each initialization. To this end, we introduce OmniGlimpse, an asymmetric propose-verify framework that separates adaptive association from stable identity consolidation. At the representation level, short-term local and cross-view evidence is mapped into target-specific adaptive spaces to propose associations, whereas long-term identity distributions are maintained for verification in a canonical space defined by a frozen encoder. At the update level, short-term evidence may guide current associations, whereas, after bootstrapping, only verified observations may update persistent memory or drive adaptation. Under a unified target-specified evaluation protocol, OmniGlimpse improves over the best evaluated online baseline for each metric by 9.1 IDF1 and 22.1 CVIDF1 points on M3Track. The corresponding gains on MvMHAT are 20.3 and 26.2 points. Despite using no offline scene-specific training, it outperforms offline-trained references in IDF1 and remains competitive in cross-view association. These results support an approach to online identity maintenance that adapts to changing appearances while consolidating reliable evidence into stable identity anchors.
Oct 27, 2026

The Batch Size Can Be One: Test-Time Adaptation from Classifier Geometry

Test-time adaptation updates a deployed model on an unlabeled test stream. In practice this stream often arrives one sample at a time, with the classes in no particular order. A single sample then provides limited information for estimating distribution-level statistics, and selecting samples by confidence can admit predictions that are wrong yet confident. A corruption, however, shifts the features while leaving the decision regions of the frozen classifier unchanged. We therefore propose Origin-recentered Test-Time Adaptation (OriTTA), which constructs fixed geometric targets from the source model and adapts selected features toward them. Seen from the origin of the feature space, the classes lean toward a common direction. We choose a new origin to bring class scores closer together while keeping it near the mean feature of one synthetic prototype per class. From this origin, prototype directions spread more evenly, and each class obtains an axis and an ideal point before any test data arrives. During adaptation, OriTTA aligns selected features with their predicted class axis and penalizes excessive radius, keeps one input per predicted class in memory to limit feature changes for classes absent from the stream, and resets the model when re-classifying this memory reveals collapse. Without tuning for each setting, OriTTA achieves the highest mean accuracy at batch size 1 on CIFAR-10-C, CIFAR-100-C and ImageNet-C in all sixteen combinations of four backbones, two protocols and two stream orders. On ImageNet-C fog, recent methods collapse in up to two thirds of one hundred runs, while OriTTA never does.
Oct 26, 2026

Adapt for Planning, Not just Prediction: Optimistic Test-Time World Model Adaptation

Latent world models are typically pre-trained offline and kept frozen at test time, so planning with them can degrade under distribution shift. Test-time adaptation (TTA) can mitigate this degradation by updating the model during deployment. Unlike pre-training on fixed data, TTA for planning is closed-loop: the planner uses model rollouts to select actions, and the resulting transitions drive the next update. This raises the question: which objective should guide TTA for planning? Under a prediction-only objective, the model is fit only to the outcomes of its own actions, so it can remain wrong about the actions that would reach the goal, and low prediction error need not translate into better planning. Inspired by reward-biased maximum likelihood from adaptive control, we introduce Optimistic Test-Time Adaptation (OpTTA). Its objective has two terms, the prediction loss on observed transitions and the planner's goal-reaching cost on its candidate rollouts, both minimized over the model parameters. The planning term biases adaptation toward models that are optimistic about reaching the goal, without any additional environment interaction. Across visual and environment shifts on manipulation, reaching and navigation tasks, OpTTA raises final planning success over the frozen model by 13.5 percentage points on average, compared with 6.9 for the prediction-only objective (AdaJEPA), when only the predictor is adapted. When the encoder is adapted as well, the gains are 13.5 and 7.5 points. These experiments span three JEPA world models with different encoders and training objectives and, for temporal straightening, both CEM and gradient-based planners. Our results suggest that test-time adaptation of world models for planning benefits from an optimistic, task-aware term in addition to the prediction loss.
Oct 26, 2026

RoboDoctor: Training-Free Recursive Self-Improvement for Robot Policies

Generalist robot policies and language-model agents provide complementary capabilities, but completing a task also requires adapting their use to the consequences of physical interaction. We study execution as a locus of adaptation: feedback guides the current action and supplies evidence for revising subsequent decisions under fixed model weights. We introduce RoboDoctor, a training-free framework that connects diagnosis, grounded action, and persistent experience across three revision targets. The harness allocates control between policy and agent; intervention determines when to assist while preserving useful policy behavior; and operation refines physical methods through observed effects. Linked experience records retain the reason for a decision, the attempted method, its outcome, and conditions for reuse. On a curated archive of 54 RoboDojo-Sim task entries, RoboDoctor achieves a five-dimension mean score of 57.10, with a native success rate of 49.44%. Recorded interactions illustrate strategy changes, selective correction, and grasp revision. These findings characterize execution-time adaptation in the evaluated system; controlled trials remain necessary to isolate the contribution of cross-episode experience.
Oct 26, 2026

Test-Time Prompt-Agnostic Decomposition

Test-Time Prompt Tuning (TPT) adapts large pretrained models to unseen target distribution shifts by updating only lightweight prompt tokens. Existing TPT methods are effective, but they mainly optimize each prompt update locally. Across test-time steps, prompts updated at the current step are reused for the corresponding predictions and updates, so noisy or biased unlabeled objectives can accumulate and destabilize the prompt-update trajectory. We identify two failure modes: magnitude expansion, where updates move the prompt too far, and directional drift, where consecutive updates point in inconsistent directions. To analyze and control these failures, we propose Test-Time Prompt-Agnostic Decomposition (TPD\mathtt{TPD}), which characterizes observed prompt-update trajectories from a dynamical-systems view and decomposes base prompt updates into magnitude and direction. TPD\mathtt{TPD} computes a spectral radius to measure update expansion and decomposes update directions into persistent, oscillatory, and residual components. Building on this decomposition, Adaptive Koopman Control (AKC\mathtt{AKC}) regulates update magnitude by shrinking updates under expansive recent dynamics, while Hankel Update Router (HUR\mathtt{HUR}) refines update direction by preserving persistent components and suppressing oscillatory and residual components. Together, AKC\mathtt{AKC} and HUR\mathtt{HUR} produce a stabilized prompt update. As a prompt-agnostic framework, TPD\mathtt{TPD} can be plugged into visual and text TPT methods without modifying the backbone, prompt architecture, or adaptation loss. Experiments on 15 datasets show that TPD\mathtt{TPD} consistently improves 10 baselines, with accuracy gains of 33--8%8\% for visual prompts and over 2%2\% for text prompts.
Oct 26, 2026

Recovering Black-Box Point Cloud Models from Pose Shifts via Derivative-Free Optimization

Point cloud models are increasingly deployed through score-only black-box interfaces. Under this access model, test-time rotations outside the training distribution can substantially degrade performance. Although existing mitigations such as equivariant retraining and gradient-based adaptation are effective, they are infeasible without access to model weights or gradients. To address this limitation, we propose CanDO, a test-time canonicalization method that recovers lost performance using only forward queries. Specifically, CanDO searches for an input transformation that maximizes model confidence and uses derivative-free optimization to perform this search without gradients. We further identify three conditions for successful recovery: a training augmentation gap, an accuracy-aligned confidence surrogate, and global search over a multi-modal objective. Across multiple architectures and benchmarks, CanDO recovers 76% of the oracle gap under azimuthal misorientation and 51% under arbitrary SO(3) rotations for object part segmentation. These results show that frozen score-only point cloud models can recover from rotation-induced distribution shifts through post-deployment canonicalization alone.
Oct 24, 2026

SkillShift-Bench: Benchmarking Skill Learning under Environment Evolution for Web Agents

Skill learning enables web agents to reuse prior experience for long-horizon interactive tasks. However, learned skills are typically evaluated in closed environments, overlooking the fact that real-world web deployment involves continuous shifts in invocation context, perceptual grounding, and execution feasibility. To this end, we introduce SkillShift-Bench, the first benchmark for evaluating skill learning with controlled open-environment shifts. SkillShift-Bench simulates Contextual Shift, Perceptual Shift, and Execution Shift across three self-hosted web systems, and evaluates agents with diagnostics covering task success, interaction cost, skill use, and skill failure. Comprehensive experiments reveal systematic robustness degradation under controlled shifts, characterized by lower success rates, higher interaction costs, and more frequent skill failures, with Execution Shift exerting the strongest pressure. We further formulate shifted skill execution as a nonstationary local executability problem and provide both a dynamic-regret analysis and a practical test-time adaptation method, demonstrating the feasibility of improving learned-skill execution without relearning the skill library. Code, tasks, environments, documentation, and the leaderboard are available at https://skillshift-bench.github.io/.
Oct 23, 2026

KINESYS: Test-Time Adaptation for Reference-Based Effect Editing in Images

Visual effects are often easier to demonstrate via reference than to describe in words. We study effect-reference editing, where a reference image shows a desired visual effect and language identifies which properties should guide the edit. Existing single-pass editing techniques often fail to make fine-grained edits, resulting in unintended changes to source content or deviations from the reference effect. We introduce KINESYS, a test-time adaptation framework for effect-reference editing that combines reference-aware and source-aware feedback to transfer the desired effect while preserving the composition and content of the original shot. To ensure KINESYS can produce a diverse set of edits that preserve the source, we develop a robust optimization strategy in which the generative model is fine-tuned across multiple noise samples and optimization runs. Evaluation with VLM judges, source preservation metrics (LPIPS, TPIPS), and a user study shows that KINESYS outperforms state-of-the-art image editors in source preservation; on effect transfer, it remains competitive with Nano Banana 2 and outperforms open-weight editors.
Oct 23, 2026

Interactive image diagnosis using test-time backpropagation from the auxiliary task

Multi-task deep learning has been developed to ensure explainability in computer-aided diagnosis (CAD) systems. These systems simultaneously predict interpretable characteristics of findings in auxiliary tasks to help radiologists assess the reliability of the diagnostic prediction. However, the CAD's opinion would not be referred to when the auxiliary task prediction deviates from the radiologist's expectation. To further leverage CAD's knowledge in the above situation, we proposed a multi-task gradient assist (MGA) method that brings interactivity to multi-task models without additional training. MGA allows radiologists to give feedback to the CAD and ask for additional diagnostic suggestions. Using two datasets of mammography and CT, we show that MGA successfully improves diagnostic accuracy upon feedback on various auxiliary tasks. Compared to an existing method using forward propagation only, MGA performs more robustly and can be combined to further improve accuracy. Our method paves a promising way for integrating artificial intelligence into image diagnosis.
Oct 23, 2026

Can Test-Time Memory Evolution Outperform Test-Time Training for Scientific Discovery?

Recent test-time training (TTT) frameworks have enabled open-weight large language models (LLMs) to outperform inference-only frameworks powered by frontier models on scientific discovery tasks. However, TTT introduces additional computational cost for online weight updates during the discovery process, making computational efficiency an important consideration when scaling to larger models. This motivates us to explore an efficient test-time adaptation method for scientific discovery. We introduce Test-Time Memory Evolution (TTME), a plug-and-play test-time adaptation method, as an alternative to TTT. TTME continually updates the external memory by distilling the experience to guide subsequent discovery without updating model weights. To have a clean comparison, we evaluate TTME and TTT under the same discovery framework, replacing only the test-time adaptation method. Our evaluation shows that TTME outperforms TTT on all five mathematical tasks, while reducing computational cost by 82.882.8-84.6%84.6\% in FLOPs and 47.747.7-58.4%58.4\% in runtime. These results establish TTME as an effective and computationally efficient alternative to TTT in the evaluated discovery settings.
Oct 21, 2026

No Frames, Just Frequencies: Bandwidth-Efficient Deployment-Time Adaptation for Smart Glasses

Deep neural networks deployed on smart glasses must adapt to changes in the distribution of incoming data while adhering to strict privacy and device constraints. We present No Frames, Just Frequencies, a bandwidth-efficient source-assisted adaptation framework that characterizes distribution shifts using compact frequency descriptors computed on-device. Rather than transmitting camera frames, the glasses compute and send frequency descriptors to a host device that contains a replica of the model. The host device maps the received frequency descriptor to a distribution shift operator that synthesizes out-of-distribution data and adapts a model replica using a pre-defined adaptation protocol. We evaluate the framework on mixtures of ImageNet-C corruptions using convolution-based (ConvNeXtV2-Femto), self-attention-based (DeiT-Tiny), and hybrid (TinyViT-5M) vision models, on person classification (Visual Wake Words), and on object detection (MS-COCO). Our results reveal that adaptation driven solely by the transmitted descriptors recovers 93−95%93-95\% of the supervised oracle adaptation gain on all three architectures, 82%82\% on Visual Wake Words and 9494--98%98\% on MS-COCO. Notably, adaptation quality degrades gracefully as the descriptor resolution decreases from 224224 to 2828 pixels, consistently across architectures, enabling bandwidth reductions from 213×213\times to 784×784\times across the frequency descriptors.