Programs and reusable artifacts
Neural Programmer-Interpreters
Learns neural programs that share a recurrent interpreter and a program memory.
First arXiv submission 2015-11-19 · ICLR 2016
Reading list
Papers
Research on generated weights, neural compilers, and fuzzy-function programming.
Browse by topic or first arXiv submission date, or download the bibliography.
Showing 13 papers
No papers match these filters. Try a broader term or reset the filters.
Generated weights and adapters
HyperNetworks
Uses one neural network to generate the weights of another.
First arXiv submission 2016-09-27 · ICLR 2017
Parameter-efficient building blocks
Prefix-Tuning: Optimizing Continuous Prompts for Generation
Adapts a frozen language model by learning a short sequence of continuous prefix vectors.
First arXiv submission 2021-01-01 · ACL-IJCNLP 2021
Generated weights and adapters
Learning to Generate Task-Specific Adapters from Task Description
Introduces Hypter, which generates adapters from natural-language task descriptions.
First arXiv submission 2021-01-02 · ACL-IJCNLP 2021
Parameter-efficient building blocks
LoRA: Low-Rank Adaptation of Large Language Models
Adapts a pretrained model with low-rank weight updates while keeping the original weights fixed.
First arXiv submission 2021-06-17 · ICLR 2022
Generated weights and adapters
HyperTuning: Toward Adapting Large Language Models without Back-propagation
Produces task-specific parameters for a frozen language model from a few demonstrations.
First arXiv submission 2022-11-22 · ICML 2023
Generated weights and adapters
HINT: Hypernetwork Instruction Tuning for Efficient Zero- & Few-Shot Generalisation
Turns instructions and examples into reusable prefixes, adapters, and encoded instruction representations.
First arXiv submission 2022-12-20 · ACL 2023
Context compression
Learning to Compress Prompts with Gist Tokens
Compresses an instruction into a few reusable token activations.
First arXiv submission 2023-04-17 · NeurIPS 2023
Programs and reusable artifacts
The ALCHEmist: Automated Labeling 500x CHEaper Than LLM Data Annotators
Asks a language model to write labeling programs that can be saved and run locally.
First arXiv submission 2024-06-25 · NeurIPS 2024
Generated weights and adapters
Generative Adapter: Contextualizing Language Models in Parameters with A Single Forward Pass
Converts new context into low-rank model updates in one forward pass.
First arXiv submission 2024-11-08 · ICLR 2025
Generated weights and adapters
Text-to-LoRA: Instant Transformer Adaption
Uses a hypernetwork to generate LoRA adapters from task descriptions.
First arXiv submission 2025-06-06 · ICML 2025
Programs and reusable artifacts
Program-as-Weights: A Programming Paradigm for Fuzzy Functions
Proposes fuzzy-function programming: compile a natural-language function specification into a neural program, then reuse it locally.
First arXiv submission 2026-07-02 · NeurIPS 2026 (accepted)
Programs and reusable artifacts
Compile by Training: Turning Natural-Language Specifications into Local Neural Functions
Compiles a function description into a reusable local program by generating training examples and optimizing an adapter.
First arXiv submission 2026-09-03 · EMNLP 2026 System Demonstrations