AWhat changed
Introduced BitLinear-based Transformer training with binary forward-pass weights, investigating scaling and inference-memory/computation savings from training low-bit models from scratch.
Scientific scope
Introduced BitLinear-based Transformer training with binary forward-pass weights, investigating scaling and inference-memory/computation savings from training low-bit models from scratch. Not post-training quantization and not a claim that optimizer state, activations, all tensors or training arithmetic use one bit. Binary neural-network ideas are prior work; performance/energy comparisons are source-specific.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Not post-training quantization and not a claim that optimizer state, activations, all tensors or training arithmetic use one bit. Binary neural-network ideas are prior work; performance/energy comparisons are source-specific.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
- BitNet: Scaling 1-bit Transformers for Large Language Models
- Wang et al.: BitNet train-from-scratch binary-weight Transformers — official project source
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction