·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
Siri won’t be your AI girlfriend1h◆Cheaper, faster, and culturally aware, Avataar’s video AI is built for India’s scale3h◆LoHoSearch: Benchmarking Long-Horizon Search Agents Beyond the Human Difficulty Ceiling4h◆What Uncertainties Do We Need for Dynamical Systems?4h◆APPO: Agentic Procedural Policy Optimization4h◆CCKS: Consensus-based Communication and Knowledge Sharing4h◆ALIGNBEAM : Inference-Time Alignment Transfer via Cross-Vocabulary Logit Mixing4h◆Generalizing Beyond Suboptimality: Offline Reinforcement Learning Learns Effective Scheduling through Random Solutions4h◆Bootstrapped Monitoring: Leveraging Transparent Reasoning to Oversee Stronger AI Agents4h◆Can Factual Opinions Be Edited (Manipulated) in Large Language Models?4h◆Grounding Computer Use Agents on Human Demonstrations4h◆Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models4h◆LSTM based IoT Device Identification4h◆Breaking the Ice: Analyzing Cold Start Latency in vLLM4h◆From Benchmarks to Skills: Low-Rank Factors for LLM Evaluation4h◆A Controlled Study of Decoding-Time Truthfulness Methods on Instruction-Tuned LLMs4h◆Causal Inference with Generative Artificial Intelligence: Application to Texts as Treatments4h◆One Token to Fool LLM-as-a-Judge4h◆Eigenism: Ethics for a Human-AI Future4h◆GeoNatureAgent Benchmark: Benchmarking LLM Agents for Environmental Geospatial Analysis Across Frontier and Open-Weight Foundation Models4h◆Siri won’t be your AI girlfriend1h◆Cheaper, faster, and culturally aware, Avataar’s video AI is built for India’s scale3h◆LoHoSearch: Benchmarking Long-Horizon Search Agents Beyond the Human Difficulty Ceiling4h◆What Uncertainties Do We Need for Dynamical Systems?4h◆APPO: Agentic Procedural Policy Optimization4h◆CCKS: Consensus-based Communication and Knowledge Sharing4h◆ALIGNBEAM : Inference-Time Alignment Transfer via Cross-Vocabulary Logit Mixing4h◆Generalizing Beyond Suboptimality: Offline Reinforcement Learning Learns Effective Scheduling through Random Solutions4h◆Bootstrapped Monitoring: Leveraging Transparent Reasoning to Oversee Stronger AI Agents4h◆Can Factual Opinions Be Edited (Manipulated) in Large Language Models?4h◆Grounding Computer Use Agents on Human Demonstrations4h◆Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models4h◆LSTM based IoT Device Identification4h◆Breaking the Ice: Analyzing Cold Start Latency in vLLM4h◆From Benchmarks to Skills: Low-Rank Factors for LLM Evaluation4h◆A Controlled Study of Decoding-Time Truthfulness Methods on Instruction-Tuned LLMs4h◆Causal Inference with Generative Artificial Intelligence: Application to Texts as Treatments4h◆One Token to Fool LLM-as-a-Judge4h◆Eigenism: Ethics for a Human-AI Future4h◆GeoNatureAgent Benchmark: Benchmarking LLM Agents for Environmental Geospatial Analysis Across Frontier and Open-Weight Foundation Models4h◆
News/ALIGNBEAM : Inference-Time Alignment Transfer via Cross-Vocabulary Logit Mixing
arxiv
PublishedJune 12, 2026 at 4:00 AM
—neutral

ALIGNBEAM : Inference-Time Alignment Transfer via Cross-Vocabulary Logit Mixing

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

arXiv:2606.12342v1 Announce Type: cross Abstract: Domain fine-tuning degrades the safety of large language models: fine-tuned specialists readily comply with harmful prompts framed in domain language. Existing inference-time defenses that mix logits from a safe anchor model require both models to sh

Stay posted· Newsletter

A 5-min weekly brief — top movers, price watch, story of the week.

// no spam · unsubscribe one-click · free forever

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
arxivLoHoSearch: Benchmarking Long-Horizon Search Agents Beyond the Human Difficulty Ceiling4harxivWhat Uncertainties Do We Need for Dynamical Systems?4harxivAPPO: Agentic Procedural Policy Optimization4harxivCCKS: Consensus-based Communication and Knowledge Sharing4h
The Bubble Brief
WEEKLY

Read AI insights every Tuesday — top movers, new releases, story of the week.

// no spam · unsubscribe one-click · free forever

Originally published on arxiv ↗
HomeModelsNews