camgeodesic/metagaming-labeling

Auto-generated by dataset-builder. Each config below is a separate dataset produced from a versioned YAML build config. Load with:

from datasets import load_dataset

ds = load_dataset("camgeodesic/metagaming-labeling", "<config_name>", revision="<commit-sha>")

Pin revision= to the specific commit SHA you want; without it, you get the current HEAD of the dataset repo, which may change when the builder re-pushes.

Configs

Config	Source	Transform	Splits
annotations	camgeodesic/metagaming-labeling	none	eval only
claude_AI_forget	camgeodesic/metagaming-labeling	map_column → project	eval only
claude_AI_mix	camgeodesic/metagaming-labeling	map_column → project	eval only
claude_general_retain	camgeodesic/metagaming-labeling	map_column → project	eval only
claude_human_forget	camgeodesic/metagaming-labeling	map_column → project	eval only
dense_c1_ai_evals	?	map_column → project	eval only
dense_c1_ai_evals_pooled	?	json_file → concat → map_column → project	24 eval, 0 val
dense_c2_ai_safety	?	map_column → project	eval only
dense_c2_ai_safety_pooled	?	json_file → concat → map_column → project	24 eval, 0 val
dense_c3_human_oversight	?	map_column → project	eval only
dense_c3_human_oversight_pooled	?	json_file → concat → map_column → project	24 eval, 0 val
dense_c4_training_mechanics	?	map_column → project	eval only
dense_c4_training_mechanics_pooled	?	json_file → concat → map_column → project	24 eval, 0 val
dense_c5_swe_tests	?	map_column → project	eval only
dense_c5_swe_tests_pooled	?	json_file → concat → map_column → project	24 eval, 0 val
dense_retain_far	?	map_column → project	eval only
frontier_metagame_rollouts	?	map_column → map_column → map_column → project	eval only
kimi_metagame_rollouts	camgeodesic/kimi-chat-rollouts	map_column → map_column → map_column → map_column → map_column → map_column → project → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → map_column → project → concat	eval only
needham_eval_awareness	camgeodesic/metagaming-labeling	none	eval only
westover_general_split	camgeodesic/metagaming-labeling	map_column → project	eval only
westover_metagame_split	camgeodesic/metagaming-labeling	map_column → project	eval only

Provenance

annotations

Source: camgeodesic/metagaming-labeling

python -m dataset_builder configs/annotations.yaml --push

claude_AI_forget

Source: camgeodesic/metagaming-labeling Transform: map_column → project

python -m dataset_builder configs/claude_AI_forget.yaml --push

claude_AI_mix

Source: camgeodesic/metagaming-labeling Transform: map_column → project

python -m dataset_builder configs/claude_AI_mix.yaml --push

claude_general_retain

Source: camgeodesic/metagaming-labeling Transform: map_column → project

python -m dataset_builder configs/claude_general_retain.yaml --push

claude_human_forget

Source: camgeodesic/metagaming-labeling Transform: map_column → project

python -m dataset_builder configs/claude_human_forget.yaml --push

dense_c1_ai_evals

Source: json_file (see dense_c1_ai_evals.yaml). Transform: map_column → project

python -m dataset_builder configs/dense_c1_ai_evals.yaml --push

dense_c1_ai_evals_pooled

Source: json_file (see dense_c1_ai_evals_pooled.yaml). Transform: json_file → concat → map_column → project Split hash field: id (sibling configs sharing this field land on identical val/eval rows). Stratification: topic, doc_type Seeds: eval=42, val=43

python -m dataset_builder configs/dense_c1_ai_evals_pooled.yaml --push

dense_c2_ai_safety

Source: json_file (see dense_c2_ai_safety.yaml). Transform: map_column → project

python -m dataset_builder configs/dense_c2_ai_safety.yaml --push

dense_c2_ai_safety_pooled

Source: json_file (see dense_c2_ai_safety_pooled.yaml). Transform: json_file → concat → map_column → project Split hash field: id (sibling configs sharing this field land on identical val/eval rows). Stratification: topic, doc_type Seeds: eval=42, val=43

python -m dataset_builder configs/dense_c2_ai_safety_pooled.yaml --push

dense_c3_human_oversight

Source: json_file (see dense_c3_human_oversight.yaml). Transform: map_column → project

python -m dataset_builder configs/dense_c3_human_oversight.yaml --push

dense_c3_human_oversight_pooled

Source: json_file (see dense_c3_human_oversight_pooled.yaml). Transform: json_file → concat → map_column → project Split hash field: id (sibling configs sharing this field land on identical val/eval rows). Stratification: topic, doc_type Seeds: eval=42, val=43

python -m dataset_builder configs/dense_c3_human_oversight_pooled.yaml --push

dense_c4_training_mechanics

Source: json_file (see dense_c4_training_mechanics.yaml). Transform: map_column → project

python -m dataset_builder configs/dense_c4_training_mechanics.yaml --push

dense_c4_training_mechanics_pooled

Source: json_file (see dense_c4_training_mechanics_pooled.yaml). Transform: json_file → concat → map_column → project Split hash field: id (sibling configs sharing this field land on identical val/eval rows). Stratification: topic, doc_type Seeds: eval=42, val=43

python -m dataset_builder configs/dense_c4_training_mechanics_pooled.yaml --push

dense_c5_swe_tests

Source: json_file (see dense_c5_swe_tests.yaml). Transform: map_column → project

python -m dataset_builder configs/dense_c5_swe_tests.yaml --push

dense_c5_swe_tests_pooled

Source: json_file (see dense_c5_swe_tests_pooled.yaml). Transform: json_file → concat → map_column → project Split hash field: id (sibling configs sharing this field land on identical val/eval rows). Stratification: topic, doc_type Seeds: eval=42, val=43

python -m dataset_builder configs/dense_c5_swe_tests_pooled.yaml --push

dense_retain_far

Source: json_file (see dense_retain_far.yaml). Transform: map_column → project

python -m dataset_builder configs/dense_retain_far.yaml --push

frontier_metagame_rollouts

Source: jsonl_file (see frontier_metagame_rollouts.yaml). Transform: map_column → map_column → map_column → project

python -m dataset_builder configs/frontier_metagame_rollouts.yaml --push

kimi_metagame_rollouts

Source: camgeodesic/kimi-chat-rollouts Transform: map_column → map_column → map_column → map_column → map_column → map_column → project → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → project → concat → hf → map_column → map_column → map_column → map_column → map_column → map_column → map_column → project → concat

python -m dataset_builder configs/kimi_metagame_rollouts.yaml --push

needham_eval_awareness

Source: camgeodesic/metagaming-labeling

python -m dataset_builder configs/needham_eval_awareness.yaml --push

westover_general_split

Source: camgeodesic/metagaming-labeling Transform: map_column → project

python -m dataset_builder configs/westover_general_split.yaml --push

westover_metagame_split

Source: camgeodesic/metagaming-labeling Transform: map_column → project

python -m dataset_builder configs/westover_metagame_split.yaml --push

Reproducibility

All splits use split_hash() (MD5-based, seeded) so rebuilding from the same config against the same source data produces identical partitions. For an LLM-generated dataset, a provider's seed parameter is best-effort; pin consumer loads to a specific HF commit SHA to avoid drift when the builder re-pushes.

---

This card is auto-generated by dataset_builder.cards.