AI Research Radar
Friday, May 29: no frontier papers out of 341 read.
Frontier
Papers rated as moving the field
How we chooseNone of the 30 cleared papers that day was rated frontier.
This week in research
May 23–29: 145 cleared, 10 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
30 papers
- Hijacking Agent Memory: Stealthy Trojan Attacks Through Conversational InteractionTencent
- Turning Bias into Bugs: Bandit-Guided Style Manipulation Attacks on LLM JudgesCode
- Review Arcade: On the Human Alignment and Gameability of LLM ReviewsCode
- PokerSkill: LLMs Can Play Expert-Level Poker without Training or SolversCode
- SMolLM: Small Language Models Learn Small Molecular Grammar
- Stop Suppressing the Tail: Causal Inference for Extreme Events
- RightNow-Arabic-0.5B-Turbo: An Open Sub-1B Arabic Language Model via Vocabulary Injection and Edge-First Deployment
- Benchmarking Open-Source Safety Guard Models: A Comprehensive Evaluation
- The Biosecurity Blind Spot: Systematic Dual-use Detection in Open Science Infrastructure
- Measuring Real-World Prompt Injection Attacks in LLM-based Resume Screening
- FormInv: A Measurement Protocol for Semantic Invariance in Mathematical Reasoning Benchmarks
- When Models Disagree: Rethinking LLM Evaluation for Public Comment Analysis
All 30 papers
- ReasonBreak: Probing Vulnerabilities in Reasoning-Enabled Vision-Language-Action Models for Autonomous Driving
- PRO-CUA: Process-Reward Optimization for Computer Use Agents
- Anytime-Valid Federated Conformal RAG for LLM Swarms
- Relevance as a Vulnerability: How Web Retrieval Degrades Safety Alignment in LLM Agents
- Rethinking Literature Search Evaluation: Deep Research Helps, and Human Citation Lists Are Not a Ground Truth
- Provably Secure Agent Guardrail
- Harmless Yet Harmful: Neutral Prompting Attacks for Stealthy Hallucination Steering in Agent Skills
- SciIntBench: Measuring LLM Compliance with Research Integrity Norms Under Adversarial Framing
- KBF: Knowledge Boundary as Fingerprint for Language Model and Black-Box API Auditing
- Training Deliberative Monitors for Black-Box Scheming Detection
- PhAIL: A Real-Robot VLA Benchmark and Distributional Methodology
- Evolutionary Dynamics of Cooperation in Next-Generation LLM Agent Systems: A Cross-Provider Empirical Extension
- Label Over Logic? How Source Cues Bias Human Fallacy Judgments More Than LLMs
- Fingerprinting Inference Systems of Large Language Models
- Token Inflation: How Dishonest Providers Can Overcharge for Large Language Model Usage
- Dissociative Identity: Language Model Agents Lack Grounding for Reputation Mechanisms
- Resolution Diagnostics for Paired LLM Evaluation
- GPIC: A Giant Permissive Image Corpus for Visual Generation