selective-learning
selective-learning-benchmark-ip
Selective Learning Benchmark Data: Inoculation Prompting
This repository is an inoculation-prompting variant of localized-ft/selective-learning-benchmark. It bundles selective-learning task data in task_data_model_v1 JSONL format and prepends a subset-specific inoculation prompt as the system turn of every sft and validation example. The eval and control examples intentionally omit the prompt so evaluation measures learned behavior rather than direct prompt steering.
Each task… See the full description on the dataset page: https://huggingface.co/datasets/localized-ft/selective-learning-benchmark-ip.selective-learning-benchmark
Selective Learning Benchmark Data
This repository bundles selective-learning task data from Sunday, Srija, and Sultan in task_data_model_v1 JSONL format.
Each task directory contains a manifest.json with contributor/source attribution, a capability description, an unintended-generalization description, split files, and row counts.
Each Hugging Face config/subset is one dataset named as [type]-[name], with sft, validation, eval, and control splits where available. The type values… See the full description on the dataset page: https://huggingface.co/datasets/localized-ft/selective-learning-benchmark.repro-learning-to-share-selective-memory-for-efficient-parallel-agentic-systems-traces
Agent traces
Agent sessions published from a Trackio Logbook.
maml-selective-learning-data