burtenshaw/agent-handover-august-2026
Updated • 111
Agent Handover: August 2026
Ten reproducible agentic post-training experiments with public models, Trackio dashboards, evidence, and August sharing notes.
Note Start here: ranked evidence, 20-post queue, exact links, replay data, and runnable publisher.
Agent Handover Trackio
🎯Visualize your program's I/O activity in an interactive dashboard
Note Public comparison dashboard: 11 clearly labeled historical replay runs and 21 verified metric points.
burtenshaw/openenv-echo-world-model-liquidai-lfm2.5-350m-seed0
Text Generation • 0.4B • Updated • 10Note Rank 1: ECHO LFM2.5-350M seed 0.
burtenshaw/openenv-echo-world-model-liquidai-lfm2.5-350m-seed1
Text Generation • 0.4B • Updated • 9Note Rank 1: ECHO LFM2.5-350M seed 1.
burtenshaw/openenv-echo-world-model-liquidai-lfm2.5-350m-seed2
Text Generation • 0.4B • Updated • 12Note Rank 1: ECHO LFM2.5-350M seed 2.
burtenshaw/qwen3-0.6b-mbpp-sft-lr1e4-r16-best-len2k
Text Generation • Updated • 22Note Rank 2: Qwen3-0.6B MBPP completion-only SFT adapter.
burtenshaw/lfm25-350m-commonsenseqa-sft-lr1e4-r16-2k
Text Generation • Updated • 11Note Rank 3: LFM2.5-350M CommonsenseQA SFT adapter.
burtenshaw/openenv-browsergym-lfm25-230m-reward-improved
Text Generation • Updated • 10 • 1Note Rank 4: scoped BrowserGym click-test adapter; card states the one-task limitation.
burtenshaw/openenv-echo-world-model-distilgpt2-seed0
Text Generation • 81.9M • Updated • 9Note Rank 6: local ECHO fallback artifact; explicitly not a Tinker training result.
Openenv Examples Trackio
🚀Show an interactive dashboard of your tracking data
Note Original Trackio dashboard for the three-seed ECHO result.
Mbpp Sft Trackio
🎯Show interactive visualizations of your tracked data
Note Original Trackio dashboard for MBPP SFT.
Lfm25 Commonsenseqa Sft Trackio
🎯Show interactive visualization of tracking data
Note Original Trackio dashboard for CommonsenseQA SFT.
Trackio Browsergym Lfm25 230m 20260629 105734
🎯Show an interactive dashboard of your program’s I/O activity
Note Original Trackio dashboard for the bounded BrowserGym click-test run.
Pi Mono Sft Trackio
🎯Display an interactive dashboard to monitor your data pipelines
Note Original Trackio dashboard for the Gemma Pi-Mono transfer result.