This checkpoint was trained with examples/echo_world_model/train_echo.py to predict OpenEnv terminal environment outputs from verifier-free ECHO loss.

Training Metrics

metric	value
`best_step`	`10`
`heldout_ce_after`	`0.41339555382728577`
`heldout_ce_before`	`13.741455078125`
`heldout_ce_delta`	`-13.328059524297714`
`heldout_ce_improvement_pct`	`96.9916173252615`
`heldout_token_acc_after`	`0.8571428571428571`
`heldout_token_acc_before`	`0.0`
`lr`	`5e-05`
`model`	`LiquidAI/LFM2.5-350M`
`seed`	`0`
`steps`	`60`

burtenshaw/openenv-echo-world-model-liquidai-lfm2.5-350m-seed0

OpenEnv ECHO World Model

Training Metrics

burtenshaw/
openenv-echo-world-model-liquidai-lfm2.5-350m-seed0