132,259 rows. Concatenation of:
DCAgent/g1_min_episodes_e1_gpt_long_top8_glm47_traces (37,925 rows) — top8 + extra long swegym traces
DCAgent/g1_diverse_tezos_top4_100k_glm47_traces (94,334 rows) — balanced 4-way mix of swesmith/issue/superuser/tezos
No deduplication; shuffled with seed=42. Schema: intersection columns (agent, conversations, date, episode, model, model_provider, result, run_id, task, trial_name).… See the full description on the dataset page:
https://huggingface.co/datasets/DCAgent2/g1_gptlong_plus_diverse_tezos_glm47_traces.