AI Research Radar
April 28 – May 4: no frontier papers out of 570 read.
Frontier
Papers rated as moving the field
How we chooseNone of the 16 cleared papers this week was rated frontier.
The week by field
Apr 28 – May 4: 16 cleared, 0 frontier
The week’s papers have not been sorted into fields yet.
Models under test
Papers this week that name the model
Labs publishing
Papers this week with an author at the lab
Also cleared
16 papers
- RAFP: Identifying LLM Lineages via Rare-Region FingerprintsMeta
- Logic Jailbreak: Efficiently Unlocking LLM Safety Restrictions Through Formal Logical ExpressionBaidu
- EVADE-Bench: Multimodal Benchmark for Evaluating and Enhancing Evasive Content DetectionAlibaba
- Beyond Black-Box Obfuscation: Mechanistic Analysis and Defense of White-Box MonitorsCode
- Ranking Free RAG: Replacing Re-ranking with Selection in RAG for Sensitive DomainsCode
- Secure LLM Fine-Tuning via Safety-Aware ProbingCode
- Beyond Public Access in LLM Pre-Training Data
- Harmonia: End-to-End RAG Serving Optimization
- SHIELD: Safety on Humanoids via CBFs In Expectation on Learned Dynamics
- The Accountability Paradox: How Platform API Restrictions Undermine AI Transparency Mandates
- Extracting memorized pieces of (copyrighted) books from open-weight language models
- Dynamic Optimization and Safety Indicator Injection for Jailbreaking Text-to-Image Models with Multimodal Safety Filters