·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
India becomes the latest country to report rogue AI agent activity51m◆An Nvidia insider-turned-VC on why robotics hasn’t hit its ChatGPT moment, yet2h◆Physical AI is more than robots, says the Nvidia veteran now investing in it2h◆Elorian’s first AI model wants to reason with images, not just about them2h◆Jev is Vercel’s fastest-adopted AI model ever3h◆A week with RTX Spark: everything we’ve learned using Nvidia’s new laptops3h◆Mark Wahlberg on using AI in film ‘the right way’ and being open to an AI clone carrying on his legacy4h◆Meta’s new terms may allow it to train AI on news sites for free4h◆Veritas expands its anti-counterfeit chip to jewelry, trophies, and fragrance bottles4h◆Replit’s CEO says the future of work is the ‘self-driving company’5h◆Optimal Hydra: Model Heterogeneous Federated Learning for LLMs using Data-Informed Hydra-Nets14h◆Judging the Judges: Confidence and Disagreement in Rubrics as Rewards14h◆Content associations with answer selection and engagement in supply chain community question answering14h◆ORACLE-SWE: Quantifying the Contribution of Critical Information Signals on SWE Agents14h◆REKI: Registering Knowledge in VLAs14h◆Multimodal Remote Inference14h◆HyCo: Hybrid Drafting for Cost-Efficient Mobile-Cloud Speculative Decoding14h◆Arithmetic Circuits and Neural Networks for Regular Matroids14h◆Linear Independence of Polynomial Compositions with Bounded Input Degree14h◆Training Reasoning Models to Generate Diverse Scientific Ideas14h◆India becomes the latest country to report rogue AI agent activity51m◆An Nvidia insider-turned-VC on why robotics hasn’t hit its ChatGPT moment, yet2h◆Physical AI is more than robots, says the Nvidia veteran now investing in it2h◆Elorian’s first AI model wants to reason with images, not just about them2h◆Jev is Vercel’s fastest-adopted AI model ever3h◆A week with RTX Spark: everything we’ve learned using Nvidia’s new laptops3h◆Mark Wahlberg on using AI in film ‘the right way’ and being open to an AI clone carrying on his legacy4h◆Meta’s new terms may allow it to train AI on news sites for free4h◆Veritas expands its anti-counterfeit chip to jewelry, trophies, and fragrance bottles4h◆Replit’s CEO says the future of work is the ‘self-driving company’5h◆Optimal Hydra: Model Heterogeneous Federated Learning for LLMs using Data-Informed Hydra-Nets14h◆Judging the Judges: Confidence and Disagreement in Rubrics as Rewards14h◆Content associations with answer selection and engagement in supply chain community question answering14h◆ORACLE-SWE: Quantifying the Contribution of Critical Information Signals on SWE Agents14h◆REKI: Registering Knowledge in VLAs14h◆Multimodal Remote Inference14h◆HyCo: Hybrid Drafting for Cost-Efficient Mobile-Cloud Speculative Decoding14h◆Arithmetic Circuits and Neural Networks for Regular Matroids14h◆Linear Independence of Polynomial Compositions with Bounded Input Degree14h◆Training Reasoning Models to Generate Diverse Scientific Ideas14h◆
News/Will the Judge Flip? Predicting Position-Sensitive LLM Judgments from Residual Stream Activations
arxiv
PublishedOctober 8, 2026 at 4:00 AM

Will the Judge Flip? Predicting Position-Sensitive LLM Judgments from Residual Stream Activations

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

arXiv:2610.07115v1 Announce Type: cross Abstract: The order in which candidate responses are presented can change an LLM judge's verdict. Detecting such a position flip ordinarily requires judging each pair in both orders, which doubles the number of judgments. We investigate whether residual stream

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
Source
↗
arxiv
Read original ↗All from arxiv →

No replies yet. Be first.

Source
↗
arxiv
Read original ↗All from arxiv →

Related coverage

More from ARXIV
arxivOptimal Hydra: Model Heterogeneous Federated Learning for LLMs using Data-Informed Hydra-Nets14harxivJudging the Judges: Confidence and Disagreement in Rubrics as Rewards14harxivContent associations with answer selection and engagement in supply chain community question answering14harxivORACLE-SWE: Quantifying the Contribution of Critical Information Signals on SWE Agents14h
The Bubble Brief
MON · WED · FRI

Read AI 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