·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
NewsBench: Expert-Grounded Evaluation of Epistemic Quality in AI News Reporting5h◆AdaPRA: Process Reward Agents with Adaptive Rubrics for Data Science5h◆Scoped Semantic Invariance as a Foundation for LLM Reliability5h◆An actively trained neural operator for near-explicit atmospheric chemistry5h◆InferPriv: Evaluating Inferential Privacy Leakage in LLM Assistants5h◆ThinkBit: Search Codewords, Recalibrate Scales, Select by Reasoning5h◆Multimodal Requirements Traceability Link Recovery in Industrial Settings5h◆The Price of Scale in Proactive Memory Recall5h◆GUIFlow: Temporal Redundancy-Aware Incremental Inference for GUI Agents5h◆The Value of Memory in Language Agents5h◆HyScale: Scaling Long-Context Training for Learned Sparse Attention5h◆Training an MoE LLM in MXFP4: How We Closed the Gap to BF165h◆WiMoT: Replay-Free Continual Wi-Fi Sensing via Moment Transport5h◆Advancing SVD-based LLM Compression via Layer-Wise Error Model Search5h◆Unbiased Alignment for Large Language Models with Noisy Preferences5h◆AdaptTok-Indic: Script-Aware Adaptive Tokenization for Indic Languages5h◆Synthetic Child-Adult Interactions Are Not Developmentally Plausible5h◆VX-One: Typed, Calibrated Decision Heads for Real-Time Perimeter Triage5h◆Conditional Flow Models for Offline Maximum Entropy Reinforcement Learning5h◆Neural Disjunctive Normal Forms for Learning Exact Rank Filters5h◆NewsBench: Expert-Grounded Evaluation of Epistemic Quality in AI News Reporting5h◆AdaPRA: Process Reward Agents with Adaptive Rubrics for Data Science5h◆Scoped Semantic Invariance as a Foundation for LLM Reliability5h◆An actively trained neural operator for near-explicit atmospheric chemistry5h◆InferPriv: Evaluating Inferential Privacy Leakage in LLM Assistants5h◆ThinkBit: Search Codewords, Recalibrate Scales, Select by Reasoning5h◆Multimodal Requirements Traceability Link Recovery in Industrial Settings5h◆The Price of Scale in Proactive Memory Recall5h◆GUIFlow: Temporal Redundancy-Aware Incremental Inference for GUI Agents5h◆The Value of Memory in Language Agents5h◆HyScale: Scaling Long-Context Training for Learned Sparse Attention5h◆Training an MoE LLM in MXFP4: How We Closed the Gap to BF165h◆WiMoT: Replay-Free Continual Wi-Fi Sensing via Moment Transport5h◆Advancing SVD-based LLM Compression via Layer-Wise Error Model Search5h◆Unbiased Alignment for Large Language Models with Noisy Preferences5h◆AdaptTok-Indic: Script-Aware Adaptive Tokenization for Indic Languages5h◆Synthetic Child-Adult Interactions Are Not Developmentally Plausible5h◆VX-One: Typed, Calibrated Decision Heads for Real-Time Perimeter Triage5h◆Conditional Flow Models for Offline Maximum Entropy Reinforcement Learning5h◆Neural Disjunctive Normal Forms for Learning Exact Rank Filters5h◆
News/Adapting in the Dark: Efficient and Stable Test-Time Adaptation for Black-Box Models
arxiv
PublishedApril 20, 2026 at 4:00 AM
▲bullish

Adapting in the Dark: Efficient and Stable Test-Time Adaptation for Black-Box Models

Source
arxiv.orgfull article ↗
Read on arxiv→
Publisher summary· verbatim

arXiv:2604.15609v1 Announce Type: new Abstract: Test-Time Adaptation (TTA) for black-box models accessible only via APIs remains a largely unexplored challenge. Existing approaches such as post-hoc output refinement offer limited adaptive capacity, while Zeroth-Order Optimization (ZOO) enables input

Models mentioned
02
  • 01google logo
    vit-base-patch16-224-in21k
    google/vit-base-patch16-224-in21k
  • 02openai logo
    clip-vit-base-patch16
    openai/clip-vit-base-patch16
Compare these 2 models→
Related
02
  • arxivJul 10
    Efficient Long-Horizon Learning for Learned Optimization
  • arxivApr 10
    Information as Structural Alignment: A Dynamical Theory of Continual Learning
Stay posted· Newsletter

Three short emails a week: Monday, Wednesday and Friday. Top movers, new models and the most-referenced story.

// no spam · unsubscribe one-click · free forever

Your email is used only to send The Bubble Brief. Privacy policy

Discussion
Mentioned models
06
  • 01
    BETA
  • 02
    vit-base-patch16-224-in21k
    google/vit-base-patch16-224-in21k
  • 03
    clip-vit-base-patch16
    openai/clip-vit-base-patch16
  • 04
    TENT
  • 05
    TPT
  • 06
    ZOO
Source
↗
arxiv
Read original ↗All from arxiv →
Tags
04
#machine-learning#computer-vision#test-time-adaptation#real-time-inference
Mentioned companies
03
DataCiteGoogleOpenAI

No replies yet. Be first.

Mentioned models
06
  • 01
    BETA
  • 02
    vit-base-patch16-224-in21k
    google/vit-base-patch16-224-in21k
  • 03
    clip-vit-base-patch16
    openai/clip-vit-base-patch16
  • 04
    TENT
  • 05
    TPT
  • 06
    ZOO
Source
↗
arxiv
Read original ↗All from arxiv →
Tags
04
#machine-learning#computer-vision#test-time-adaptation#real-time-inference
Mentioned companies
03
DataCiteGoogleOpenAI

Related coverage

More from ARXIV
arxivNewsBench: Expert-Grounded Evaluation of Epistemic Quality in AI News Reporting5harxivAdaPRA: Process Reward Agents with Adaptive Rubrics for Data Science5harxivScoped Semantic Invariance as a Foundation for LLM Reliability5harxivAn actively trained neural operator for near-explicit atmospheric chemistry5h
The Bubble Brief
MON · WED · FRI

Read machine-learning insights every Monday, Wednesday and Friday — top movers, new models, story of the week.

// no spam · unsubscribe one-click · free forever

Your email is used only to send The Bubble Brief. Privacy policy

Originally published on arxiv ↗
PrivacyTermsRefundsLegal notice|AboutMethodologyContact

© 2026 Databubble

Built by Marouane Gazouzi · Readyo, AI interview coach

HomeModelsNews