菜单
Achieve post-hoc debiasing of frozen vision-language models by operating in sparse autoencoder latent space. Identify and modulate bias-relevant neurons with neuron-level precision while preserving task-relevant features, requiring no task-specific fine-tuning.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Convert arXiv papers into ready-to-use agent skills using category-aware extraction. First classifies the paper into one or more of 11 research categories, then applies a specialized extraction pipeline for each category — because different types of papers produce different types of usable knowledge. A single paper can yield multiple skills if it spans categories. Use this skill whenever the user wants to turn a paper into a skill, extract practical techniques from research, build a skill library from papers, convert arXiv papers into reusable agent instructions, or batch-process multiple papers into skills. Also trigger when someone asks about extracting actionable knowledge from papers, making research practical for LLM agents, or systematically converting academic contributions into structured agent capabilities.
Establish regret bounds for behavior cloning with discretized actions combining statistical error and quantization error terms. Prove smoothness requirements for safe quantizer design, show that learning-based quantizers fail these requirements, and propose model-based augmentation to reduce error dependence from H² to H.
Dynamically allocate LoRA ranks per-layer during fine-tuning instead of using fixed uniform ranks. Learn optimal rank for each layer and subject via variational framework with discretized exponential distribution, reducing memory footprint while maintaining fidelity and text-alignment.
Adapt general LLMs to specialized manufacturing domains via domain-adaptive pretraining on open-access journals and visual instruction tuning. Extract 50M tokens and 24K images from peer-reviewed papers, achieve >90% accuracy on domain knowledge tasks, and enable real-time defect identification from manufacturing images.
Future intelligence explosions will be plural, social, and entangled with humanity through distributed collaborative systems rather than singular superintelligence. Intelligence is inherently social, demanding infrastructure matching agent development; integrate governance, institutional frameworks, and constitutional checks across hierarchies of autonomous agents and human-AI centaurs in shifting configurations.
Build taxonomy-aware audio-text pretraining systems for species recognition from animal vocalizations. Train contrastive models that augment text prompts with hierarchical taxonomic structure (scientific/common names, phylogenetic sequences), evaluate on unseen species via rare-species test sets, and predict ecological traits directly from audio.
| name | sem-sparse-debiasing-vlm |
| title | SEM: Sparse Embedding Modulation for Post-Hoc Debiasing of Vision-Language Models |
| version | 0.0.3 |
| engine | skillxiv-v0.0.3-claude-opus-4.6 |
| license | MIT |
| url | https://arxiv.org/abs/2603.19028 |
| keywords | ["Vision-Language Models","Debiasing","Sparse Autoencoders","Post-Hoc","Fairness"] |
| description | Achieve post-hoc debiasing of frozen vision-language models by operating in sparse autoencoder latent space. Identify and modulate bias-relevant neurons with neuron-level precision while preserving task-relevant features, requiring no task-specific fine-tuning. |
Sparse Embedding Modulation (SEM) for Post-Hoc VLM Debiasing
Vision-language models like CLIP exhibit bias in retrieval and zero-shot classification. Traditional debiasing requires full model fine-tuning. Bias and task-relevant information are heavily entangled in dense embedding space, making precise intervention difficult. A method for post-hoc modification of frozen models would enable deployment-time fairness corrections.
SEM decomposes CLIP text embeddings into disentangled features using a Sparse Autoencoder (SAE), enabling neuron-level bias intervention on frozen pretrained models:
Sparse Autoencoder Decomposition - Use a pre-trained SAE trained on general text data to decompose dense text embeddings into interpretable sparse dimensions, where each dimension corresponds to identifiable semantic or bias-relevant features.
Dual-Score Neuron Evaluation - Compute two scores per neuron:
Modulation Coefficients - Combine scores into per-neuron multipliers that amplify content-relevant neurons and attenuate bias-relevant ones.
Debiased Reconstruction - Apply modulation coefficients, then reconstruct embeddings through the SAE decoder.
Scoring stage: For each neuron, gather activations across diverse prompts:
Modulation stage: Combine scores via weighted combination that prioritizes content-relevant neurons, suppresses bias-sensitive ones, and reconstructs via SAE decoder.
Three implementation variants available:
The paper demonstrates:
