Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges
Paper • 2604.13602 • Published • 19
AI & ML interests
Our long-term goal is to achieve AGI by creating AI scientists. Specifically, our research directions include deeply unlocking intelligence through LLM, providing a favorable intelligent environment through AI4S foundational models, as well as exploring the in-depth interaction between these two aspects.
Recent Activity
View all activity
fdu-lcz
submitted a
paper to Daily Papers about 11 hours ago
kyqiu
authored 3
papers about 1 month ago
Piloting Structure-Based Drug Design via Modality-Specific Optimal Schedule
Paper • 2505.07286 • Published
MolCRAFT: Structure-Based Drug Design in Continuous Parameter Space
Paper • 2404.12141 • Published
Empower Structure-Based Molecule Optimization with Gradient Guided Bayesian Flow Networks
Paper • 2411.13280 • Published
thomaswang
updated a
model about 2 months ago
thomaswang
published a
model about 2 months ago
fdu-lcz
authored 2
papers 3 months ago
fdu-lcz
submitted 2
papers to Daily Papers 3 months ago
fdu-lcz
authored 8
papers 3 months ago
AMix-1: A Pathway to Test-Time Scalable Protein Foundation Model
Paper • 2507.08920 • Published
Multi-Programming Language Sandbox for LLMs
Paper • 2410.23074 • Published
Searching for Best Practices in Retrieval-Augmented Generation
Paper • 2407.01219 • Published • 11
Aligning Large Language Models with Human Preferences through Representation Engineering
Paper • 2312.15997 • Published • 2
Explainable Synthetic Image Detection through Diffusion Timestep Ensembling
Paper • 2503.06201 • Published
Benchmark^2: Systematic Evaluation of LLM Benchmarks
Paper • 2601.03986 • Published • 34
Controllable Memory Usage: Balancing Anchoring and Innovation in Long-Term Human-Agent Interaction
Paper • 2601.05107 • Published • 24
RECAST: Expanding the Boundaries of LLMs' Complex Instruction Following with Multi-Constraint Data
Paper • 2505.19030 • Published • 1
hhhhhhh789
authored 2
papers 5 months ago