·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
Nvidia, Microsoft launch open AI security alliance – without OpenAI, Google, or Anthropic1h◆The path to artificial superintelligence1h◆Closing the data loop in AI-driven drug discovery1h◆Building the enterprise environment for agentic AI1h◆NVIDIA Cosmos-H-Dreams: Bringing Real-Time Generative Simulation to Surgical Robotics3h◆A Consensus-Based Framework for Relative Preference Evaluation of Large Language Models9h◆Probing Latent Colombian Identity Inferences in Qwen2.5-7B with Natural Language Autoencoders9h◆Data Quality over Capacity: Internalizing Documents into LoRA Adapters for Closed-Book QA9h◆Enjoy Your Talk: A Human-Centered Benchmark for Multi-Turn Dialogue with Decoupled User Simulation, Target Modeling, and Judging9h◆Multi-Mask Diffusion Language Models for Few-Step Generation9h◆Solar Open 2 Technical Report9h◆The Geometry of Personality: Activation Steering with Jungian Cognitive Functions9h◆Self-Guided Process Reward Optimization with Redefined Step-wise Advantage for Process Reinforcement Learning9h◆H$^2$SD: Hybrid Hindsight Self-Distillation9h◆LunarFM: A Shared Multimodal Representation of the Moon's Surface9h◆Prior laundering: learned priors with inherited, undetectable overconfidence9h◆Deep Sigma Point Processes for RCS Modeling in Spaceborne SAR Imagery9h◆Prompt as a Data Type: In-Database LLM Prompt Management and Rewriting9h◆CausalForge: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal Inference9h◆Quantum Spectral Model: Data Reuploading with Input-Conditioned Frequency Support9h◆Nvidia, Microsoft launch open AI security alliance – without OpenAI, Google, or Anthropic1h◆The path to artificial superintelligence1h◆Closing the data loop in AI-driven drug discovery1h◆Building the enterprise environment for agentic AI1h◆NVIDIA Cosmos-H-Dreams: Bringing Real-Time Generative Simulation to Surgical Robotics3h◆A Consensus-Based Framework for Relative Preference Evaluation of Large Language Models9h◆Probing Latent Colombian Identity Inferences in Qwen2.5-7B with Natural Language Autoencoders9h◆Data Quality over Capacity: Internalizing Documents into LoRA Adapters for Closed-Book QA9h◆Enjoy Your Talk: A Human-Centered Benchmark for Multi-Turn Dialogue with Decoupled User Simulation, Target Modeling, and Judging9h◆Multi-Mask Diffusion Language Models for Few-Step Generation9h◆Solar Open 2 Technical Report9h◆The Geometry of Personality: Activation Steering with Jungian Cognitive Functions9h◆Self-Guided Process Reward Optimization with Redefined Step-wise Advantage for Process Reinforcement Learning9h◆H$^2$SD: Hybrid Hindsight Self-Distillation9h◆LunarFM: A Shared Multimodal Representation of the Moon's Surface9h◆Prior laundering: learned priors with inherited, undetectable overconfidence9h◆Deep Sigma Point Processes for RCS Modeling in Spaceborne SAR Imagery9h◆Prompt as a Data Type: In-Database LLM Prompt Management and Rewriting9h◆CausalForge: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal Inference9h◆Quantum Spectral Model: Data Reuploading with Input-Conditioned Frequency Support9h◆
News/Reproduction Beyond Benchmarks: ConstBERT and ColBERT-v2 Across Backends and Query Distributions
arxiv
PublishedApril 18, 2026 at 4:00 AM
▼bearish

Reproduction Beyond Benchmarks: ConstBERT and ColBERT-v2 Across Backends and Query Distributions

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

arXiv:2604.09982v2 Announce Type: replace-cross Abstract: Reproducibility must validate architectural robustness, not just numerical accuracy. We evaluate ColBERT-v2 and ConstBERT across five dimensions, finding that while ConstBERT reproduces within 0.05% MRR@10 on MS-MARCO, both models show a drop

Stay posted· Newsletter

A 5-min weekly brief — top movers, price watch, story of the week.

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

Discussion
Mentioned models
02
  • 01
    ColBERT-v2
  • 02
    ConstBERT
Source
↗
arxiv
Read original ↗All from arxiv →
Tags
04
#information retrieval#reproducibility#evaluation#architectural robustness

No replies yet. Be first.

Mentioned models
02
  • 01
    ColBERT-v2
  • 02
    ConstBERT
Source
↗
arxiv
Read original ↗All from arxiv →
Tags
04
#information retrieval#reproducibility#evaluation#architectural robustness

Related coverage

More from ARXIV
arxivA Consensus-Based Framework for Relative Preference Evaluation of Large Language Models9harxivProbing Latent Colombian Identity Inferences in Qwen2.5-7B with Natural Language Autoencoders9harxivData Quality over Capacity: Internalizing Documents into LoRA Adapters for Closed-Book QA9harxivEnjoy Your Talk: A Human-Centered Benchmark for Multi-Turn Dialogue with Decoupled User Simulation, Target Modeling, and Judging9h
The Bubble Brief
WEEKLY

Read information retrieval insights every Tuesday — top movers, new releases, story of the week.

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

Originally published on arxiv ↗
HomeModelsNews