LoRA Research Atlas — An Interactive Map of Low-Rank Adaptation

LoRA Research Atlas is an interactive literature map of low-rank adaptation. It connects papers by research direction and publication time, explains where methods overlap or explicitly extend earlier work, and brings their experimental results into a searchable archive.

Explore the Research Tree Browse Experimental Results

LoRA Research Atlas showing the relationship between HiRA and ABBA

Explore the research

  • 65 papers, nine research directions — follow the development of parameterization, rank allocation, subspaces, optimization, quantization, composition, generalization, reinforcement learning, and continual learning.
  • Explained relationships — distinguish chronological order from explicit extensions and shared mechanisms. Filter connections, focus on a method, or open a relationship to read its explanation and original evidence.
  • Original papers one click away — each paper node links directly to its source. Zoom, pan, search, and replay the research timeline.

Inspect the experiments

  • 632 sourced results across 132 model/configuration groups — browse a dense results matrix or switch to a method’s training and evaluation settings.
  • Conditions alongside numbers — inspect the base model, rank, dataset, metric, split, and source table before comparing results.
  • Search, filter, and export — filter by method, research direction, task, or model; choose dataset columns; export the selected records as CSV.

The archive contains selected experimental configurations rather than every ablation. Results retain each paper’s evaluation conditions; they are not presented as a unified leaderboard. Missing entries are left blank. The bibliography was reviewed on November 14, 2026, and the experimental archive on November 17, 2026.

Direct Link