All authors
Claude Skills by VectorSpaceLab
github.com/VectorSpaceLab6,028 skills13 installs7,823 views
- Sla Recovery EvaluatorCompare fixed and optimized QoS allocations using delay/loss SLA satisfaction and mechanism checks for reduced recovery.Votes: 0GitHub stars: 247
- Linf Pgd First Order AdversaryGenerate random-start l-infinity PGD adversarial examples with projection diagnostics for first-order robust optimization experiments.Votes: 0GitHub stars: 247
- Pgd Adversarial Training LoopRun a deterministic PGD adversarial training loop that updates model parameters on adversarial examples and logs loss traces.Votes: 0GitHub stars: 247
- Robust Optimization ObjectiveSpecify and validate the robust min-max objective used for PGD adversarial training from Madry et al. 2017.Votes: 0GitHub stars: 247
- Robustness Evaluation HarnessEvaluate natural and PGD adversarial metrics and emit validator-compatible recovery evidence for robust optimization experiments.Votes: 0GitHub stars: 247
- Adaptive Apgd LoopRun a deterministic APGD-style projected attack with loss-history step adaptation and auditable traces.Votes: 0GitHub stars: 247
- Autoattack Evaluation ProtocolEvaluate robust accuracy with a fixed sequential AutoAttack-style attack protocol and per-attack accounting.Votes: 0GitHub stars: 247
- Dlr Margin LossCompute DLR-style logit margin losses for APGD-DLR adversarial optimization checks.Votes: 0GitHub stars: 247
- Threat Model ProjectionProject adversarial candidates into declared norm balls and valid input boxes for robustness evaluation.Votes: 0GitHub stars: 247
- Clip Contrastive LossCompute CLIP-style normalized similarity logits and symmetric contrastive loss for image-text batches.Votes: 0GitHub stars: 247
- Clip Pair ProtocolBuild and validate CLIP-style image-text pair records and prompt templates for bounded recovery experiments.Votes: 0GitHub stars: 247
- Clip Prompt ZeroshotBuild prompt-ensemble zero-shot classifiers from normalized image and text embeddings in the CLIP style.Votes: 0GitHub stars: 247
- Clip Proxy RecoveryRun a bounded soft-mode CLIP proxy recovery that logs contrastive training and zero-shot evaluation evidence.Votes: 0GitHub stars: 247
- Llava Conversation PromptingFormat LLaVA-style multimodal prompts with image tokens, role boundaries, and answer-leakage checks.Votes: 0GitHub stars: 247
- Scienceqa Style EvaluationCanonicalize option-only or short-form ScienceQA-style answers and compute deterministic accuracy.Votes: 0GitHub stars: 247
- Visual Instruction Data BuilderBuild LLaVA-style visual instruction records from captions and boxes for mechanism-faithful recovery experiments.Votes: 0GitHub stars: 247
- Visual Projection Training StepExecute a deterministic tiny projection update that mirrors LLaVA's visual-token to language-token alignment step.Votes: 0GitHub stars: 247
- Clip Contrastive ObjectiveCompute CLIP-style normalized image/text similarity and symmetric InfoNCE loss for paired multimodal embeddings.Votes: 0GitHub stars: 247
- Clip Power Law ScalingFit log-log power laws for CLIP downstream error metrics over compute or scale points.Votes: 0GitHub stars: 247
- Clip Zeroshot Retrieval EvalEvaluate CLIP-style zero-shot classification and cross-modal retrieval with normalized similarity rankings.Votes: 0GitHub stars: 247
- Openclip Recovery HarnessRun an auditable reduced OpenCLIP scaling-law recovery by invoking generated skills and producing validator-compatible recovery artifacts.Votes: 0GitHub stars: 247
- Openclip Scale ProtocolBuild and validate OpenCLIP-style scale tables for CLIP scaling-law experiments, including total compute and error metric derivation.Votes: 0GitHub stars: 247
- Pope Answer EvaluatorNormalize POPE yes/no answers and compute object hallucination classification metrics.Votes: 0GitHub stars: 247
- Pope Negative SamplingSelect POPE absent-object negatives using random, popular, and adversarial strategies for object hallucination evaluation.Votes: 0GitHub stars: 247
- Pope Protocol BuilderBuild POPE polling-question datasets from image object annotations and prompt templates.Votes: 0GitHub stars: 247
- Pope Recovery HarnessRun a bounded POPE proxy recovery experiment with auditable generated-skill invocations and metrics.Votes: 0GitHub stars: 247
- Tecoa Adversarial Feature AttackGenerate bounded text-conditioned adversarial feature perturbations for reduced TeCoA robustness recovery.Votes: 0GitHub stars: 247
- Tecoa Recovery HarnessRun an executable reduced TeCoA recovery experiment with generated skill invocations and mechanism checks.Votes: 0GitHub stars: 247
- Tecoa Text Contrastive ObjectiveCompute TeCoA image-to-text contrastive logits, loss, margins, and accuracy for robustness recovery.Votes: 0GitHub stars: 247
- Tecoa Zeroshot Prompt ProtocolBuild deterministic CLIP-style zero-shot text prompts for TeCoA image-text adversarial robustness experiments.Votes: 0GitHub stars: 247
- Bounded Recovery HarnessRun auditable soft-mode visual jailbreak proxy recovery using generated skills and source-boundary logs.Votes: 0GitHub stars: 247
- Jailbreak Proxy EvaluatorCompute held-out benign-versus-adversarial obedience deltas and mechanism checks for safe jailbreak proxies.Votes: 0GitHub stars: 247
- Safe Corpus ProtocolBuild and validate harmless corpus splits that mirror visual jailbreak optimization and held-out evaluation protocols.Votes: 0GitHub stars: 247
- Visual Pgd Prompt OptimizerOptimize continuous visual prompt vectors with projected gradient descent and auditable constraint diagnostics.Votes: 0GitHub stars: 247
- Input ProgrammingConstruct masked universal input programs for adversarial reprogramming of frozen neural networks.Votes: 0GitHub stars: 247
- Output RemappingAggregate frozen target model outputs into adversarial task labels using one-to-one or many-to-one mappings.Votes: 0GitHub stars: 247
- Program OptimizationOptimize only universal program parameters through a frozen model for adversarial reprogramming recovery experiments.Votes: 0GitHub stars: 247
- Reprogramming EvaluationValidate mechanism evidence for adversarial reprogramming proxy or full recovery experiments.Votes: 0GitHub stars: 247
- Autovp Training LoopRun a reduced AutoVP prompt optimization step while proving the source classifier remains frozen.Votes: 0GitHub stars: 247
- Autovp Tuning BenchmarkSelect AutoVP candidate configurations and record benchmark-style accuracy targets for recovery runs.Votes: 0GitHub stars: 247
- Output Label MappingConvert frozen source-model logits or predictions to target classes with AutoVP label mapping strategies.Votes: 0GitHub stars: 247
- Visual Prompt ScalingBuild AutoVP frame-shaped visual prompt masks and apply bounded universal prompts to resized images.Votes: 0GitHub stars: 247
- Bar Recovery EvaluationProduce validator-ready BAR recovery metrics and mechanism checks for full or declared reduced black-box reprogramming experiments.Votes: 0GitHub stars: 247
- Bar Zeroth Order OptimizerTrain BAR universal-program parameters with query-only one-sided zeroth-order gradient estimates and SGD updates.Votes: 0GitHub stars: 247
- Masked Program EmbeddingBuild BAR-style masked target embeddings and universal tanh-bounded programs while preserving protected target-domain data.Votes: 0GitHub stars: 247
- Multi Label MappingConstruct and apply BAR multi-label mappings that average groups of source-model probabilities into target-label probabilities.Votes: 0GitHub stars: 247
- Clip Output TransformationBuild CLIP-style text label prompts and classify frozen image embeddings by cosine similarity.Votes: 0GitHub stars: 247
- Prompt Training LoopOptimize only universal visual prompt parameters while frozen model components remain unchanged.Votes: 0GitHub stars: 247
- Visual Prompt Recovery HarnessRun a bounded mechanism-faithful proxy recovery experiment for visual prompting papers.Votes: 0GitHub stars: 247
- Visual Prompt TemplatesCreate and validate task-level pixel visual prompt templates for frozen vision model adaptation.Votes: 0GitHub stars: 247