{"apiVersion":"v1","taxonomyVersion":1,"generatedAt":"2026-10-29T14:23:50.737Z","data":{"recordId":"wsa_5b9uv6iwsgbr","score":0.86,"confidence":"high","capped":false,"capReason":null,"sources":[{"n":1,"url":"https://huggingface.co/webbrain-one/lfm3-vld-2b","publisher":"Hugging Face","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-29T00:30:11.026Z","detectedPublishedAt":"2026-10-28T07:27:39.000Z","rights":"original"}],"claims":[{"id":"af6e55d2-0e49-4f67-90c0-fba3b46f6eee","kind":"capability","statement":"The model combines a 24-layer hybrid decision backbone, a vision tower, a multimodal projector, and embeddings from LiquidAI/d1-2B.","citations":[{"source":1,"supportProbability":0.97,"supports":true,"quote":{"text":"**Decision Backbone**: The complete 24-layer hybrid architecture (Conv + GQA attention + Feed-Forward), vision tower, multimodal projector, and embeddings from `LiquidAI/d1-2B` (trained on browser actions, verification, and decision trajectories).","rights":"original"}}]},{"id":"d83c1d04-1c18-4f79-b513-616dc4ec1348","kind":"capability","statement":"The model uses the original language head from LiquidAI/LFM3-VL-2B, untied from token embeddings.","citations":[{"source":1,"supportProbability":0.97,"supports":true,"quote":{"text":"**Restored Language Head**: The original uncorrupted generative language head (`lm_head.weight`) from `LiquidAI/LFM3-VL-2B`, untied from token embeddings (`tie_word_embeddings: false`).","rights":"original"}}]},{"id":"4fbd242f-a168-4c7d-ba9f-c9af8351a52f","kind":"capability","statement":"The publisher reports 91.3% accuracy for the Decision Probe on the d1 Balanced Decision Benchmark.","citations":[{"source":1,"supportProbability":0.92,"supports":true,"quote":{"text":"- **Preserved decision separability**: Hidden states retain deep decision knowledge, reaching **91.27% accuracy** with **0% false completions** on the balanced browser decision benchmark.","rights":"original"}}]},{"id":"7535e036-c84d-4c83-9b1d-79fb6cc2ed4e","kind":"capability","statement":"A Transformers quickstart uses AutoModelForImageTextToText and AutoProcessor with trust_remote_code enabled.","citations":[{"source":1,"supportProbability":0.78,"supports":true,"quote":null}]},{"id":"b85962bd-579b-4745-a8f7-11408d21a923","kind":"availability","statement":"The model is published at webbrain-one/lfm3-vld-2b on Hugging Face.","citations":[{"source":1,"supportProbability":0.36,"supports":false,"quote":{"text":"`LFM3-VLD-2B` (Restored Language Head) is a hybrid multimodal browser decision model built using the **Pegasus method** (inspired by `AveLabs/Qwen3.8-27B-Perplexing-Pegasus`).","rights":"original"}}]}]},"nextCursor":null,"license":{"catalog":{"name":"CC BY 4.0","spdx":"CC-BY-4.0","url":"https://creativecommons.org/licenses/by/4.0/","attribution":"wowsuchai","attributionUrl":"https://wowsuchai.withglint.app","covers":["catalog metadata","classifications","generated summaries","feed event metadata"]},"excluded":{"covers":["source quotations","linked source content"],"terms":"Not licensed by wowsuchai. Source quotations and linked source content keep their original rights."}}}