OneBee companion training corpora: PMB probes, behavioral probe sets, and the SFT / DPO / distillation data behind the Study 001 checkpoints.
Arjhine Ty
arjhinety
AI & ML interests
MoE, Distillation
Recent Activity
updated a collection 6 days ago
OneBee Datasets updated a collection 6 days ago
OneBee Datasets updated a collection 6 days ago
OneBee DatasetsOrganizations
Canonical training corpora for OpenGrad experiments, plus evaluation evidence for a run whose weights no longer exist.
-
arjhinety/OpenGrad-ToolPolicy-Canonical-v2
Viewer • Updated • 173k • 242 -
arjhinety/OpenGrad-ToolPolicy-Canonical-v1
Viewer • Updated • 214k • 536 -
arjhinety/OpenGrad-ToolPolicy-Canonical-v2-M0-snapshot
Viewer • Updated • 103k • 354 -
arjhinety/OpenGrad-Qwen3.5-2B-M0-SFT-CorpusV1-evaluation
Preview • Updated • 41
SchemaForge models for structured JSON extraction and schema-aware generation.
Qwen3.5-2B tool-policy post-training: every Study 001 checkpoint and corpus, negative results included. Frozen 2026-09-13.
-
arjhinety/OpenGrad-Qwen3.5-2B-M1-DPO-CanonicalV2-Final-v2
Updated -
arjhinety/OpenGrad-Qwen3.5-2B-M1-DPO-CanonicalV2-Final-v2-GGUF
Text Generation • 2B • Updated • 762 -
arjhinety/OpenGrad-Qwen3.5-2B-M0-SFT-CanonicalV2-Final
Text Generation • 2B • Updated • 397 -
arjhinety/OpenGrad-ToolPolicy-Canonical-v2
Viewer • Updated • 173k • 242
Post-training checkpoints for tool-use policy research. Each card states the measured behaviour, rejected checkpoints, and what was not measured.
OneBee multimodal companion checkpoints: SFT, DPO, distilled, and GGUF releases.
OneBee companion training corpora: PMB probes, behavioral probe sets, and the SFT / DPO / distillation data behind the Study 001 checkpoints.
Qwen3.5-2B tool-policy post-training: every Study 001 checkpoint and corpus, negative results included. Frozen 2026-09-13.
-
arjhinety/OpenGrad-Qwen3.5-2B-M1-DPO-CanonicalV2-Final-v2
Updated -
arjhinety/OpenGrad-Qwen3.5-2B-M1-DPO-CanonicalV2-Final-v2-GGUF
Text Generation • 2B • Updated • 762 -
arjhinety/OpenGrad-Qwen3.5-2B-M0-SFT-CanonicalV2-Final
Text Generation • 2B • Updated • 397 -
arjhinety/OpenGrad-ToolPolicy-Canonical-v2
Viewer • Updated • 173k • 242
Canonical training corpora for OpenGrad experiments, plus evaluation evidence for a run whose weights no longer exist.
-
arjhinety/OpenGrad-ToolPolicy-Canonical-v2
Viewer • Updated • 173k • 242 -
arjhinety/OpenGrad-ToolPolicy-Canonical-v1
Viewer • Updated • 214k • 536 -
arjhinety/OpenGrad-ToolPolicy-Canonical-v2-M0-snapshot
Viewer • Updated • 103k • 354 -
arjhinety/OpenGrad-Qwen3.5-2B-M0-SFT-CorpusV1-evaluation
Preview • Updated • 41
Post-training checkpoints for tool-use policy research. Each card states the measured behaviour, rejected checkpoints, and what was not measured.
SchemaForge models for structured JSON extraction and schema-aware generation.
OneBee multimodal companion checkpoints: SFT, DPO, distilled, and GGUF releases.