·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
Perturbation-Based Data Selection for Medical Question Answering1h◆Adaptive Transfer of Representational Signatures for LLM Reliability Detection1h◆Propasafe-Hybrid: A Text-Based Hybrid Propaganda Detection Tool1h◆Incremental Predictive Value of LLM Tutoring Interactions for Student Performance1h◆Small Edits, Shifting Decisions: Benchmarking LLM Accuracy and Reliability on French Input Variants1h◆Five Architectures, One Shared Space Cross-Model Representational Alignment Without Weight Merging1h◆Trajectory2Gym: Decision-Grounded Compositional Interaction Synthesis for Tool-Use Agents1h◆When the Weakest Language Keeps Moving: ARMOR-DP for Multilingual BabyLM1h◆EdgeDeID: A 0.6B Clinical De-identifier Trained Without Real Patient Data1h◆When Personas Become Harmful: Guiding LLMs with Global Citizenship Values Alignment1h◆"High-Quality" Data for Early-Stage Instruction Tuning: An Empirical Study of Data Quality Metrics1h◆Reversible Pseudonymisation of Documents Sent to External LLMs: an Open Bilingual Benchmark and an Engine1h◆Memory Tax: Dissecting the Net Value of Agent Memory1h◆The LALA Benchmark: Evaluating Literary and Linguistic Analysis in LLMs1h◆Higher Reasoning Effort Improves Reference Verifiability and Changes Reasoning Summaries in GPT-5.51h◆Synthetic linguistic data1h◆SAUSS: Stochastic Approximation with Unbiased Simulated Scores for Limited Dependent Variable Models1h◆A Russian, a Brazilian, and a German Walk into a Bar, Ask the Same Question-- And Get Three Different Answers: Profiling LLM Perspective Diversity on Contested Topics1h◆Transformer to GPT-5: Advancements, Limitations, and Future Directions in Large Language Model Design1h◆Factual Yet One-Sided: Stance Poisoning and Evidence Balancing in Retrieval-Augmented Generation1h◆Perturbation-Based Data Selection for Medical Question Answering1h◆Adaptive Transfer of Representational Signatures for LLM Reliability Detection1h◆Propasafe-Hybrid: A Text-Based Hybrid Propaganda Detection Tool1h◆Incremental Predictive Value of LLM Tutoring Interactions for Student Performance1h◆Small Edits, Shifting Decisions: Benchmarking LLM Accuracy and Reliability on French Input Variants1h◆Five Architectures, One Shared Space Cross-Model Representational Alignment Without Weight Merging1h◆Trajectory2Gym: Decision-Grounded Compositional Interaction Synthesis for Tool-Use Agents1h◆When the Weakest Language Keeps Moving: ARMOR-DP for Multilingual BabyLM1h◆EdgeDeID: A 0.6B Clinical De-identifier Trained Without Real Patient Data1h◆When Personas Become Harmful: Guiding LLMs with Global Citizenship Values Alignment1h◆"High-Quality" Data for Early-Stage Instruction Tuning: An Empirical Study of Data Quality Metrics1h◆Reversible Pseudonymisation of Documents Sent to External LLMs: an Open Bilingual Benchmark and an Engine1h◆Memory Tax: Dissecting the Net Value of Agent Memory1h◆The LALA Benchmark: Evaluating Literary and Linguistic Analysis in LLMs1h◆Higher Reasoning Effort Improves Reference Verifiability and Changes Reasoning Summaries in GPT-5.51h◆Synthetic linguistic data1h◆SAUSS: Stochastic Approximation with Unbiased Simulated Scores for Limited Dependent Variable Models1h◆A Russian, a Brazilian, and a German Walk into a Bar, Ask the Same Question-- And Get Three Different Answers: Profiling LLM Perspective Diversity on Contested Topics1h◆Transformer to GPT-5: Advancements, Limitations, and Future Directions in Large Language Model Design1h◆Factual Yet One-Sided: Stance Poisoning and Evidence Balancing in Retrieval-Augmented Generation1h◆
News/OpenEvoShield: Dual Non-Stationary Continual Defense for Open-World Multi-Agent System Attacks
arxiv
PublishedJuly 23, 2026 at 4:00 AM
▲bullish

OpenEvoShield: Dual Non-Stationary Continual Defense for Open-World Multi-Agent System Attacks

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

arXiv:2607.19351v1 Announce Type: new Abstract: LLM-based multi-agent systems (LLM-MAS) are increasingly deployed in safety-critical applications, where adversaries inject malicious instructions through inter-agent communication to propagate harmful behaviors. Unlike static threats, these attacks ar

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
04
  • 01
    M1
  • 02
    M2
  • 03
    M3
  • 04
    M4
Source
↗
arxiv
Read original ↗All from arxiv →
Tags
04
#safety#security#continual-learning#multi-agent-systems

No replies yet. Be first.

Mentioned models
04
  • 01
    M1
  • 02
    M2
  • 03
    M3
  • 04
    M4
Source
↗
arxiv
Read original ↗All from arxiv →
Tags
04
#safety#security#continual-learning#multi-agent-systems

Related coverage

More from ARXIV
arxivPerturbation-Based Data Selection for Medical Question Answering1harxivAdaptive Transfer of Representational Signatures for LLM Reliability Detection1harxivPropasafe-Hybrid: A Text-Based Hybrid Propaganda Detection Tool1harxivIncremental Predictive Value of LLM Tutoring Interactions for Student Performance1h
The Bubble Brief
MON · WED · FRI

Read safety 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