Researchers release TabletCraft, an open-source system for bidirectional Akkadian–English translation and cuneiform rendering
The system combines a ByT5-based NMT model, a cuneiform sign converter, and a visual tablet renderer to enable both reading ancient tablets and composing new cuneiform messages.
1 source · cross-referenced
- TabletCraft is the first open-source system enabling bidirectional interaction with Mesopotamian cuneiform, allowing users to translate Akkadian to English and compose new messages as rendered cuneiform tablets.
Researchers have released TabletCraft, an open-source toolkit that bridges a 4,000-year gap in Mesopotamian cuneiform literacy by enabling bidirectional Akkadian–English translation and cuneiform rendering. The system integrates three components: a ByT5-based neural machine translation model trained on 116,000 bidirectional samples, a cuneiform sign converter with 14,240 mappings (95.3% coverage), and a visual tablet renderer that produces rendered clay tablet images. Users can translate Akkadian inscriptions to English and, in the reverse direction, compose new English messages that are translated to Akkadian, converted to cuneiform signs, and rendered as a virtual clay tablet.
On a held-out validation split of 2,812 samples from the Akkademia dataset, the system achieves 49.1 BLEU for Akkadian-to-English translation and 48.5 BLEU for English-to-Akkadian, marking the first published quantitative results for reverse-direction translation. The toolkit is distributed as a pip-installable package with both a command-line interface and a web demo, and the authors note it is the first open-source system to enable bidirectional interaction with cuneiform writing.
- Oct 13, 2026 · Google DeepMind — Blog
Google DeepMind details safety guardrails for Gemini 4 Argon’s cyberdefense and biology capabilities
Trust76 - Oct 13, 2026 · Google DeepMind — Blog
Google DeepMind unveils Gemini 4 Argon for robotics with 87% score on Embodied Reasoning benchmark
Trust78 - Oct 13, 2026 · arXiv cs.CL
Researchers propose Elastic Latent Recurrence to unify recurrent and non-recurrent language models without inference-time trade-offs
Trust79