Papers10

#cosine similarity

DARE: Aligning LLM Agents with the R Statistical Ecosystem via Distribution-Aware Retrieval

DARE is a new way for AI assistants to find the right R functions by also looking at what the data looks like, not just the words in the question.

#distribution-aware retrieval#RPKB#RCodingAgent

Compositional Generalization Requires Linear, Orthogonal Representations in Vision Embedding Models

Intermediate

Arnas Uselis, Andrea Dittadi et al.Feb 27arXiv

The paper asks a simple question: what must a vision model’s internal pictures (embeddings) look like if it can recognize new mixes of things it already knows?

#compositional generalization#linear representation hypothesis#orthogonal representations

Reinforced Fast Weights with Next-Sequence Prediction

Intermediate

Hee Seung Hwang, Xindi Wu et al.Feb 18arXiv

Fast weight models remember context with a tiny, fixed memory, but standard next-token training teaches them to think only one word ahead.

#fast weight models#next-sequence prediction#reinforcement learning for LMs

Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines?

Intermediate

Anton Korznikov, Andrey Galichin et al.Feb 15arXiv

Sparse autoencoders (SAEs) are popular for explaining what large language models are doing, but this paper shows they often don’t learn real, meaningful features.

#sparse autoencoders#interpretability#dictionary learning

VidVec: Unlocking Video MLLM Embeddings for Video-Text Retrieval

Intermediate

Issar Tzachor, Dvir Samuel et al.Feb 8arXiv

VidVec shows that video-capable multimodal language models already hide strong matching signals between videos and sentences inside their middle layers.

#video–text retrieval#multimodal large language models#intermediate layer embeddings

Semantic Search over 9 Million Mathematical Theorems

Intermediate

Luke Alexander, Eric Leonen et al.Feb 5arXiv

This paper builds a Google-for-theorems: a semantic search engine that finds exact theorems, lemmas, and propositions instead of just entire papers.

#semantic theorem search#mathematical information retrieval#dense retrieval

CGPT: Cluster-Guided Partial Tables with LLM-Generated Supervision for Table Retrieval

Intermediate

Tsung-Hsiang Chou, Chen-Jui Yu et al.Jan 22arXiv

This paper introduces CGPT, a way to help computers find the right tables by building smarter mini-versions of tables and training with tough practice questions.

#table retrieval#synthetic query generation#cluster-guided partial tables

InfiniteVGGT: Visual Geometry Grounded Transformer for Endless Streams

Intermediate

Shuai Yuan, Yantai Yang et al.Jan 5arXiv

InfiniteVGGT is a streaming 3D vision system that can keep working forever on live video without running out of memory.

#InfiniteVGGT#rolling memory#causal attention

Relational Visual Similarity

Intermediate

Thao Nguyen, Sicheng Mo et al.Dec 8arXiv

Most image-similarity tools only notice how things look (color, shape, class) and miss deeper, human-like connections.

Enriching Word Vectors with Subword Information

Intermediate

Piotr Bojanowski, Edouard Grave et al.Jul 15arXiv

This paper teaches computers to understand words by also looking at the smaller pieces inside words, like 'un-', 'play', and '-ing'.

#subword embeddings#character n-grams#skip-gram