hypernetworks

提出 CLIK，一个上下文条件化的参数生成器，通过单次前向传播将上下文映射为 LoRA 参数，把知识内化到模型权重中，无需测试时梯度更新，在文档问答和多轮对话上优于现有超网络方法。

论文提出 WebMix，一种利用混合超网络在部署时组合已有低秩适配器（LoRA）的网页智能体适配方法，无需在目标界面上重新训练即可实现跨域和跨版本迁移，并在 TimeWarp 与 WebArena-Lite-v2 上取得显著提升。

The paper proposes three approaches for few-shot author-style adaptation of LLMs—contrastive activation steering, a steering-vector prediction network, and a hypernetwork that predicts LoRA adapters—and finds the hypernetwork offers the best trade-off between style imitation and output quality for both seen and unseen authors.

HyperThink introduces a text-to-parameter approach where a lightweight hypernetwork predicts query-conditioned updates to a small subset of an LLM's parameters, amortizing long thinking traces into a single forward pass to improve the accuracy-latency tradeoff for reasoning tasks.

A University of Amsterdam study combines RL-based Multiple Instance Learning, adversarial debiasing, and preference-conditioned hypernetworks for student-at-risk prediction, finding that fairness objectives can be added to an interpretable RL-MIL pipeline but preference conditioning alone fails to yield controllable trade-offs between performance and Equalized Odds due to mode collapse.

HyperThink 提出一种 text-to-parameter 超网络方法，通过一次前向传播直接预测 LLM 的参数更新，从而免去长思维链解码，在低延迟场景下显著提升数学与通用推理的准确率-延迟权衡。

HyperZip proposes an efficient LLM-based lossless data compression framework that uses diffusion-based LLMs with Multi-Token Prediction and a hypernetwork to adapt models to target data without fine-tuning, achieving a superior compression-rate/throughput trade-off over autoregressive baselines.

Atelier introduces a self-supervised hypernetwork framework that generates implicit neural representations for cryoEM density maps, enabling local feature extraction and improving performance on downstream annotation tasks.

HyperQ introduces token-conditioned quantum residual branches into frozen masked-diffusion language models, using a circuit hypernetwork to dynamically generate quantum circuit parameters. This approach improves performance on benchmarks and is computationally efficient, requiring fewer fine-tuning examples than classical baselines.

This paper introduces CLAW, a method that uses hypernetworks to generate low-rank adapters for world models, enabling efficient online adaptation in model-based reinforcement learning with limited test-time data.

A new paper from NaceAI proposes a hypernetwork-based method for injecting knowledge into large language models without modifying their core parameters, potentially enabling efficient continual learning. The approach uses generated low-rank adapters to encode new facts while keeping the base model frozen.

The paper investigates scaling laws for hypernetwork-based knowledge injection into LLMs, finding predictive power law scaling and reliable out-of-distribution generalization, establishing hypernetworks as a scalable alternative to LoRA and full fine-tuning.

PorTAL is a novel architecture that decouples task fine-tuning from specific base model weights, enabling portable task adapters that can be transferred to new models with minimal retraining. It achieves ~98% of LoRA's accuracy gain on unseen models using only half the calibration data.

Independent study shows 227M-parameter hypernetwork adds zero gain over well-crafted few-shot prompts for tool-use in 3B Llama, achieving 79.7% of GPT-5 performance at 10× lower latency.