All Archives - Finisky Garden

0%

Good , 125 posts in total , Keep on posting.╰(*°▽°*)╯

2026

05-12

NV-Embed: Training a Decoder-only Embedding Model with Latent Attention Pooling

05-12

ICLR 2025 Outstanding Paper: Shallow Safety Alignment Only Modifies the First Few Tokens

05-10

NeurIPS 2025 Best Paper Gated Attention: Non-linearity, Sparsity, and Attention Sink

05-09

MRL: Nest Embeddings During Training, Truncate at Inference

05-06

GritLM: A Single LLM for Both Embedding and Generation

05-06

HyDE: Retrieval via LLM-Generated Hypothetical Documents

05-03

LLM2Vec-Gen: Embeddings That Encode the Model's Answer, Not the Query

05-02

ICLR 2026 Outstanding Paper: LLMs Get Lost in Multi-Turn Conversation

04-30

Differentiable Search Index: A Brief Reading

04-29

Transformer Succinctness: The Other Side of Expressiveness

04-25

LLM2Vec: Turning Decoder LLMs into Embedding Models

04-22

Thoughts on Training Data Depletion and Quality Degradation

04-22

Claude Code: 1.6% AI, 98.4% Scaffolding

04-17

The Hivemind of Language Models

04-16

From RAG to Knowledge Compilation

04-15

Theoretical Ceiling of Vector Retrieval

04-14

Unexpected Perks of Talking to AI

04-13

How Claude Dreams: Background Memory Defragmentation

04-11

AI and Employment: A 200-Year-Old Debate

04-09

Three Evolutions of Agent Engineering