Dmitrii Krasheninnikov

4 papers · 2019–2024 · 3 venues · across top CS/AI venues

Authors are matched by name, so researchers who share a name ("Yang Liu", "Wei Wang") appear as a single profile and their paper counts, venues and achievements are combined. How identity is handled.

Achievements

Jump to papers ↓

🌍 Conference Polyglot · Bronze 🏃 Academic Marathon · Sprint 🌉 Interdisciplinary Bridge · Bronze 🗺️ Taxonomy Completionist (15) 🐣 Hot Topic Early Bird

Venues

NeurIPS (2) ICLR (1) ICML (1)

Top co-authors

David Krueger (3) Egor Krasheninnikov (1) Fabien Roger (1) Bruno Kacper Mlodozeniec (1) Rohin Shah (1) Tegan Maharaj (1) Ryan Greenblatt (1) Pieter Abbeel (1) Nikolaus Howe (1) Anca Dragan (1)

Research topics

Artificial Intelligence (4) Core AI (4) Machine Learning (3) Reinforcement Learning (2) Learning Paradigms (2) Learning Types · Machine Learning (2) Methods (2) Deep Learning (1) Learning Types · Deep Learning (1) Models (1) Optimization & Theory (1)

Keywords

reward function (2) large language model (2) reinforcement learning (2) model safety (1) inverse reinforcement learning (1) probing (1) reward hacking (1) capability elicitation (1) password-locked model (1) safety evaluation (1) hidden capability (1) deterministic policy (1) stochastic policy (1) fine tuning (1) llm alignment (1) proxy reward (1) causal entropy (1) dangerous capability (1) unhackable proxy (1) fine-tuning evaluation (1)

Papers

Implicit meta-learning may lead language models to trust more reliable sources ICML 2024 Stress-Testing Capability Elicitation With Password-Locked Models NeurIPS 2024 Defining and Characterizing Reward Gaming NeurIPS 2022 Preferences Implicit in the State of the World ICLR 2019