JevAny v0.3.0 · Model Family Release

Model family

Five LoRA SFT checkpoints are now available under the SimpleJev organization:

• JevAny-Gemma-4B
• JevAny-Qwen3.5-4B
• JevAny-Qwen3.5-4B-Direct-Token
• JevAny-Muse-Glimmer-30B
• JevAny-Qwen3.8-27B

JevAny-Qwen3.8-27B is the default and highest-accuracy release: 85.76% Transfer and 90.48% JevBench. The Qwen3.5 4B direct-token release reaches 80.95% JevBench; the pointer 4B release reaches 80.09% and supports larger choice sets.

Evaluation and data

• The README now compares JevAny with Kev-4B, Kev-27B, Jev 1.13.0, and Laya using the same Transfer and JevBench presentation.
• Machine-readable results include exact unrounded metrics, suite hashes, per-suite breakdowns, and released-checkpoint compute accounting.
• Current training scale is 1,772,725 text records and 2,180,242 labelled decisions. Only aggregate scale and broad task families are documented.
• The release family supports language and compatible native multimodal backbones.

Documentation and roadmap

• Added the JevAny technical report covering architecture, readout tradeoffs, ablations, compute, limitations, and open research directions.
• Updated training, inference, deployment, data, evaluation, demos, cases, and robotics documentation for the current model family.
• Historical demo recordings retain their original checkpoint identifiers for provenance.
• Full-parameter SFT, improved post-training, and hyperparameter-optimized variants remain planned releases.

Full Changelog: v0.2.0...v0.3.0

JevAny v0.2.0

Public release of the 27B SFT checkpoint and experimental RLCR checkpoint. Adds native multimodal decisions, RAGEN agent evaluation, no-label test-time adaptation experiments, Bedrock harnesses, symbolic decision trees, hardened media serving, and audited v2 results.

JevAny v0.1.0: SFT and RLCR

JevAny is a calibrated decision model focused on reinforcement learning and reliable confidence.

Models

SFT: https://huggingface.co/tianxinwei/JevAny-27B-SFT

RLCR: https://huggingface.co/tianxinwei/JevAny-27B-RLCR

Collection: https://huggingface.co/collections/tianxinwei/jevany-6ab2c941bcecb4d2c61d1326

Qwen/Qwen3.8-27B base weights are required separately.