Jonas Geiping

JonasGeiping

https://jonasgeiping.github.io/

AI & ML interests

Machine Learning Safety, Security and Privacy; Optimization in Deep Learning; Mathematical Optimization: Federated Learning

Recent Activity

upvoted a paper 9 days ago

Models That Know How Evaluations Are Designed Score Safer

upvoted a collection 9 days ago

🕵️🛡️ Evaluation Meta Knowledge

upvoted a paper 11 days ago

End-to-End Context Compression at Scale

View all activity

Organizations

upvoted a paper 9 days ago

Models That Know How Evaluations Are Designed Score Safer

Paper • 2605.28591 • Published 24 days ago • 9

upvoted a collection 9 days ago

🕵️🛡️ Evaluation Meta Knowledge

Collection

2026 arXiv preprint. Models fine-tuned on documents describing typical evaluation traits show safer behavior by having increased refusal rates and low • 11 items • Updated 11 days ago • 2

upvoted a paper 11 days ago

End-to-End Context Compression at Scale

Paper • 2606.09659 • Published 12 days ago • 27

upvoted 2 papers about 1 month ago

FutureSim: Replaying World Events to Evaluate Adaptive Agents

Paper • 2605.15188 • Published May 14 • 7

Multi-Stream LLMs: Unblocking Language Models with Parallel Streams of Thoughts, Inputs and Outputs

Paper • 2605.12460 • Published May 12 • 17

upvoted a paper 4 months ago

NESSiE: The Necessary Safety Benchmark -- Identifying Errors that should not Exist

Paper • 2602.16756 • Published Feb 18 • 4

upvoted a paper 6 months ago

Scaling Open-Ended Reasoning to Predict the Future

Paper • 2512.25070 • Published Dec 31, 2025 • 20

upvoted a paper 7 months ago

Teaching Pretrained Language Models to Think Deeper with Retrofitted Recurrence

Paper • 2511.07384 • Published Nov 10, 2025 • 20

upvoted a collection 7 months ago

Retrofitting Recurrence

Collection

21 items • Updated 24 days ago • 7

upvoted 2 papers 8 months ago

Efficient Parallel Samplers for Recurrent-Depth Models and Their Connection to Diffusion Language Models

Paper • 2510.14961 • Published Oct 16, 2025 • 8

Training Dynamics Impact Post-Training Quantization Robustness

Paper • 2510.06213 • Published Oct 7, 2025 • 3

upvoted a paper 9 months ago

Strategic Dishonesty Can Undermine AI Safety Evaluations of Frontier LLM

Paper • 2509.18058 • Published Sep 22, 2025 • 12

upvoted 2 papers 10 months ago

FAST: Factorizable Attention for Speeding up Transformers

Paper • 2402.07901 • Published Feb 12, 2024 • 3

DynaGuard: A Dynamic Guardrail Model With User-Defined Policies

Paper • 2509.02563 • Published Sep 2, 2025 • 21

upvoted a collection 12 months ago

answer-matching

Collection

Free-form datasets, human annotations, and sample-level model outputs for "Answer Matching Outperforms Multiple Choice for Language Model Evaluation" • 2 items • Updated Jul 3, 2025 • 2

upvoted 2 papers 12 months ago

Answer Matching Outperforms Multiple Choice for Language Model Evaluation

Paper • 2507.02856 • Published Jul 3, 2025 • 9

GPTailor: Large Language Model Pruning Through Layer Cutting and Stitching

Paper • 2506.20480 • Published Jun 25, 2025 • 7

upvoted 2 papers about 1 year ago

MORSE-500: A Programmatically Controllable Video Benchmark to Stress-Test Multimodal Reasoning

Paper • 2506.05523 • Published Jun 5, 2025 • 34

Zero-Shot Vision Encoder Grafting via LLM Surrogates

Paper • 2505.22664 • Published May 28, 2025 • 7

upvoted a paper over 1 year ago

Has My System Prompt Been Used? Large Language Model Prompt Membership Inference

Paper • 2502.09974 • Published Feb 14, 2025 • 9

Jonas Geiping

AI & ML interests

Recent Activity

Organizations

JonasGeiping's activity