random-lossy-singlestep-5000

Single-step SFT dataset for graphic design editing, generated using the Random Perturbation baseline from the Perturb-and-Invert (P&I) framework.

What's in it

Each example trains a model to predict a single tool call given the current design state and the history of tool calls applied so far. Trajectories are generated by sampling random tools with valid parameters from the cyberagent/crello template dataset.

Lossy inverses: palette/saturation inverse steps use negated deltas rather than per-element color snapshots, giving ~5–10× smaller trajectories with bounded per-channel drift.

Splits & size

Split	Examples
train	54,000
validation	6,000

Source: 5,000 Crello templates × ~12 steps each.

Schema

Field	Type	Description
id	string	{source_id}_step{i}
source_id	string	Originating template + perturbation ID
source	string	Perturbation class (random)
lossy	bool	Whether the inverse used lossy mode
step_idx	int	Index of this step in the trajectory
total_steps	int	Total steps in the trajectory
messages	list	[system, user, assistant] chat messages

Usage

from datasets import load_dataset
ds = load_dataset("monish-adobe/random-lossy-singlestep-5000")