Submissions by veryluckyxyz

Understanding Perception and Reasoning Through Model Merging

Building and better understanding vision-language models (2024)

HF smolagents computer-agent demo

Do Reasoning Models Show Better Verbalized Calibration?

Robustly identifying concepts introduced during chat fine-tuning with crosscoder

Retrieval with Learned Similarities

The Curse of Depth in Large Language Models

Looking Back at Speculative Decoding

Long-Context GRPO

HippoRAG: Neurobiologically Inspired Long-Term Memory for LLMs (2024)

Learning to Plan and Reason for Evaluation with Thinking-LLM-as-a-Judge

Process Reinforcement Through Implicit Rewards

Explaining Large Language Models Decisions Using Shapley Values

Phi-4 Technical Report

Alignment Faking in LLMs [pdf]

What Makes Rotary Positional Encodings Useful?

Rethinking Softmax: Self-Attention with Polynomial Activations

Post-Training Layer Scaling Prevents Forgetting and Enhances Model Merging

Random Matrix Theory in Machine Learning Tutorial

Rerankers: A Lightweight Python Library to Unify Ranking Methods

Double Descent Demystified

Synthetic Continued Pretraining

Bright: A Realistic and Challenging Benchmark for Reasoning-Intensive Retrieval

Artificial needles to real haystacks: Improving retrieval capabilities in LLMs

From Decoding to Meta-Generation: (LLMs)

Warp: On the Benefits of Weight Averaged Rewarded Policies

Experiments in Weak-to-Strong Generalization

NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models

A Case Study in CUDA Kernel Fusion