AI Research Radar

Thursday, April 30: 1 frontier paper out of 160 read.

1 frontier12 cleared160 papers read

Frontier

Papers rated as moving the field

Uncertainty-Aware Reward Discounting for Mitigating Reward Hacking

“We demonstrate statistical significance of these improvements and robustness under up to 30% supervisory noise, albeit with a trade-off in peak observed reward compared to unconstrained baselines.”

The authors, in the abstract

arXiv PDF

This week in research

Apr 24–30: 85 cleared, 6 frontier

The week’s papers have not been sorted into fields yet.

Models under test

Papers this week that name the model

Claude Sonnet 4.52
GPT-5.42
GPT-52
GPT-4o2
Gemini 3 Pro1
Llama 3.1 8B Instruct1
Mistral Small 3.21
Qwen3 8B1

Labs publishing

Papers this week with an author at the lab

Google4
Microsoft3
NVIDIA3
Alibaba1
AllenAI1
Anthropic1
IBM1
Mistral1

Also cleared

11 papers

AMMA: A Multi-Chiplet Memory-Centric Architecture for Low-Latency 1M Context Attention ServingNVIDIA
Quantamination: Dynamic Quantization Leaks Your Data Across the BatchGoogle
Woosh: A Sound Effects Foundation ModelCode
Graph Propagated Projection Unlearning: A Unified Framework for Vision and Audio Discriminative Models
Risk Reporting for Developers' Internal AI Model Use
One Word at a Time: Incremental Completion Decomposition Breaks LLM Safety
Hierarchical Long-Term Semantic Memory for LinkedIn's Hiring Agent
SafeReview: Defending LLM-based Review Systems Against Adversarial Hidden Prompts
Benchmarking the Safety of Large Language Models for Robotic Health Attendant Control
From Black-Box Confidence to Measurable Trust in Clinical AI: A Framework for Evidence, Supervision, and Staged Autonomy
Resume-ing Control: (Mis)Perceptions of Agency Around GenAI Use in Recruiting Workflows

SIGNIFICANT ON THE BOARDS

MODEL TRACKER2,648 new benchmark results across 174 models this week DEAL FLOW$158B disclosed AI funding this week across 61 rounds POLICY125 policy events tracked across 7 days