RtaForge/Anvaya-Rabbit-2.7B

Remove Anavya- filenames after Anvaya- copy

Rename Anavya- checkpoints to Anvaya- (copy)

Model card: frame as Frontier AI Research Lab (hero + About)

Model card: 0.72TG vs 0.71TG evals (benchmarks + loop rates)

0.72TG-beta: dedicated math-instruct pass (best-lineage GSM8K 0.086, general capability held, tool-call loop rate down); ship .pt + .safetensors

Add 0.72TG-beta (Phase 4 math-instruct): best-ever GSM8K, tool-call loop rate halved

Add 0.72TG-beta (Phase 4 math-instruct): best-ever GSM8K, tool-call loop rate halved

Rewrite model card: visitor-facing layout, full checkpoint eval matrix (incl. 0.6TG)

Add 0.71TG-beta: unified bridge SFT (tool-call+polish+STEM), math-tax fix verified

Publish 0.71TG-beta model card: unified bridge SFT results, math-tax fix, honest eval read

Add GAPA Phase 2 Math final checkpoint (~step 197216) as 0.7TG-beta

Remove imprint/ persona-delta checkpoints (tied to deprecated pre-TG bases)

Move pre-Tungsten (non-TG) base checkpoints to deprecated/

Add GAPA Phase 1 Logic final checkpoint (pearl_step_122032) as 0.6TG-beta

Link Anvaya Model License Agreement

Add Anvaya Model License Agreement

Add 0.5TG-beta base checkpoint (Tungsten, Siege-processed)

Update model card for 0.5TG-beta release

Resize rabbit logo to 25%

Add rabbit logo

Fix YAML metadata + trim card

fix: update weights path to Anvaya-Rabbit-2.7B-0.6-base.pt, drop safetensors row

rename weights to Anvaya-Rabbit-2.7B-0.6-base.pt

Remove misplaced base/config.json

Add rabbit_0.6 config

Add rabbit_0.6 base weights (Gå På Phase 1 logic, step 210K)

Remove v0.6/config.json

Remove v0.6/rabbit_0.6.pt

Remove v0.6/rabbit_0.6.safetensors

Add v0.6 PyTorch weights

Add v0.6 safetensors weights

Update model card for v0.6

Add v0.6 config

fix: add pipeline_tag and discoverability metadata

feat: 0.55 base — wiki warmup step 700, hash 0f85b420e694

docs: v0.55 — wiki warmup complete, checkpoint deprecation notice

Add 0.55 base: wiki warmup from 0.1-alpha, pearl step 669 hash 0f85b420e694

Restore 0.5-alpha base (bf16, post-Gurukul pretraining, pre-instruct)

restore: README to v0.5-alpha model card

restore: v0.5-alpha imprint from EFS

Revert to v0.1-alpha baseline and withdraw v0.5-alpha

Delete instruct/Anvaya-Rabbit-2.7B-0.5-alpha-instruct.pt with huggingface_hub

Delete imprint/Anvaya-Rabbit-2.7B-0.5-alpha-imprint.pt with huggingface_hub

Delete base/Anvaya-Rabbit-2.7B-0.5-alpha-base.pt with huggingface_hub

docs: corrected quickstart + v0.5-alpha maturity note

chore: add modeling_rabbit.py (safety-scrubbed AutoModel wrapper)

chore: add configuration_rabbit.py (safety-scrubbed AutoModel wrapper)

chore: add config.json (safety-scrubbed AutoModel wrapper)

docs: restore full model card (pipeline overwrote README)

Upload README.md with huggingface_hub