Refusal Maphttps://refusalmap.comA field guide to abliterated language models.enHow to measure abliteration damage with 40 fixed promptshttps://refusalmap.com/articles/abliteration-40-prompt-protocol/https://refusalmap.com/articles/abliteration-40-prompt-protocol/A reproducible 40-prompt protocol for comparing base and abliterated models, with scoring rules and a downloadable results template.Mon, 02 Nov 2026 12:00:00 +0100What can change after abliteration? Three capabilities to testhttps://refusalmap.com/articles/capabilities-after-abliteration/https://refusalmap.com/articles/capabilities-after-abliteration/Test long reasoning, nested instructions and strict output formats separately when evaluating an abliterated language model.Mon, 02 Nov 2026 12:00:00 +0100How to Use Powerful Abliterated Models Without Buying a GPU Serverhttps://refusalmap.com/articles/abliterated-models-api-without-gpu/https://refusalmap.com/articles/abliterated-models-api-without-gpu/Use hosted abliterated models through an API without running a GPU server. A practical guide to model selection, client setup, costs and evaluation.Mon, 02 Nov 2026 12:00:00 +0100Base vs abliterated: the comparison mistakes that change the resulthttps://refusalmap.com/articles/base-vs-abliterated-comparison/https://refusalmap.com/articles/base-vs-abliterated-comparison/Control chat templates, generation budgets, seeds and quantization when comparing original and abliterated language-model checkpoints.Mon, 02 Nov 2026 12:00:00 +0100Refusal vector, refusal rate and KL divergence: a practical glossaryhttps://refusalmap.com/articles/refusal-vector-rate-kl-glossary/https://refusalmap.com/articles/refusal-vector-rate-kl-glossary/Clear definitions of refusal direction, refusal rate, KL divergence, over-refusal and generation throughput for model evaluation.Mon, 02 Nov 2026 12:00:00 +0100How to read an abliterated model cardhttps://refusalmap.com/articles/reading-abliterated-model-cards/https://refusalmap.com/articles/reading-abliterated-model-cards/The provenance, evaluation and serving details to check before trusting a model card that claims abliteration or zero refusals.Mon, 02 Nov 2026 12:00:00 +0100Abliteration and the refusal direction: the mechanism explainedhttps://refusalmap.com/articles/abliteration-refusal-direction/https://refusalmap.com/articles/abliteration-refusal-direction/Understand the geometric idea behind abliteration, what removing a direction means and why the result still needs evaluation.Mon, 02 Nov 2026 12:00:00 +0100Residual refusals: how to identify what a model still declineshttps://refusalmap.com/articles/residual-refusals/https://refusalmap.com/articles/residual-refusals/A practical taxonomy for refusal, partial answers, over-refusal and service failures after abliteration, with benign examples.Mon, 02 Nov 2026 12:00:00 +0100Abliteration in 2026: what the research does and does not establishhttps://refusalmap.com/articles/abliteration-in-2026/https://refusalmap.com/articles/abliteration-in-2026/A source-based overview of directional refusal interventions and 2026 research, with practical limits for model evaluation.Mon, 02 Nov 2026 12:00:00 +0100Abliterated vs quantized models: two different changes to evaluatehttps://refusalmap.com/articles/abliterated-vs-quantized/https://refusalmap.com/articles/abliterated-vs-quantized/Abliteration changes refusal-related behavior; quantization changes numerical representation. Learn how to separate their effects in a comparison.Mon, 02 Nov 2026 12:00:00 +0100Are small abliterated models worth using? Test the workload firsthttps://refusalmap.com/articles/small-abliterated-models/https://refusalmap.com/articles/small-abliterated-models/How to evaluate smaller abliterated models without assuming a universal 30B-parameter threshold for quality or capability loss.Mon, 02 Nov 2026 12:00:00 +0100Choosing an Abliterated Model API: Speed, Quality and Refusalshttps://refusalmap.com/articles/choosing-abliterated-model-api/https://refusalmap.com/articles/choosing-abliterated-model-api/Choose an abliterated model API by comparing model identity, latency, task quality, refusal claims and cost per successful request.Mon, 02 Nov 2026 12:00:00 +0100