11 23

Lukas Galke Poech

lgalke

https://lgalke.github.io

AI & ML interests

LLM interpretability, agentic/multi-agent safety

Recent Activity

authored a paper 10 days ago

The Arbiter Agent: Continually Monitoring Multi-Agent Conversations to Detect Emergent Misalignment

upvoted a paper 11 days ago

The Arbiter Agent: Continually Monitoring Multi-Agent Conversations to Detect Emergent Misalignment

authored a paper 16 days ago

BrainSurgery: Reproducible and Reliable Declarative Weight Manipulations for Model Editing and Upcycling

View all activity

Organizations

upvoted a paper 11 days ago

The Arbiter Agent: Continually Monitoring Multi-Agent Conversations to Detect Emergent Misalignment

Paper • 2606.10747 • Published 18 days ago • 13

upvoted a paper 16 days ago

BrainSurgery: Reproducible and Reliable Declarative Weight Manipulations for Model Editing and Upcycling

Paper • 2606.09707 • Published 18 days ago • 8

upvoted a paper 18 days ago

Emergent Languages in Populations of Language Model Agents: From Token Efficiency to Oversight Evasion

Paper • 2605.31170 • Published 29 days ago • 12

upvoted a paper 22 days ago

LLMs Can Leak Training Data But Do They Want To? A Propensity-Aware Evaluation of Memorization in LLMs

Paper • 2606.06286 • Published 23 days ago • 8

upvoted a paper about 1 month ago

Confidence and Calibration of Activation Oracles for Reliable Interpretation of Language Model Internals

Paper • 2605.26045 • Published May 25 • 12

upvoted a collection 3 months ago

Activation Oracles

Collection

12 items • Updated Dec 26, 2025 • 20

upvoted a paper 8 months ago

Guarded Query Routing for Large Language Models

Paper • 2505.14524 • Published May 20, 2025 • 2

upvoted a paper 11 months ago

Dynaword: From One-shot to Continuously Developed Datasets

Paper • 2508.02271 • Published Aug 4, 2025 • 15

upvoted an article over 1 year ago

Article

Finally, a Replacement for BERT: Introducing ModernBERT

bwarner, NohTow, bclavie, orionweller, ohallstrom, staghado, alexisgallagher, rbiswasfc, fladhak, tomaarsen, ncoop57, griffin, jph00, johnowhitaker, iacolippo

•

Dec 19, 2024

• 748

upvoted a paper over 1 year ago

What makes a language easy to deep-learn? Deep neural networks and humans similarly benefit from compositional structure

Paper • 2302.12239 • Published Feb 23, 2023 • 1

upvoted an article over 1 year ago

Article

Fine-tuning LLMs to 1.58bit: extreme quantization made easy

medmekk, marcsun13, lvwerra, pcuenq, osanseviero, thomwolf

•

Sep 18, 2024

• 281

Lukas Galke Poech

AI & ML interests

Recent Activity

Organizations

lgalke's activity

Finally, a Replacement for BERT: Introducing ModernBERT

Fine-tuning LLMs to 1.58bit: extreme quantization made easy