Google Research faces backlash over TurboQuant attribution

The TurboQuant-RaBitQ spat is a classic case of institutional "inspiration" vs. independent "invention" that plagues modern AI research.

• –Critics allege Google's 8x speedup claims were boosted by comparing GPU-accelerated TurboQuant against a single-core CPU version of RaBitQ.
• –The core technical mechanism—random rotation followed by quantization—is functionally identical to RaBitQ, which was published earlier.
• –Moving primary citations to the appendix while claiming novel theoretical derivations is a common but controversial tactic for high-prestige labs.
• –Independent benchmarks for TurboQuant show significant memory savings, but the 8x performance leap remains hard to replicate without the authors' custom kernels.
• –This incident underscores the need for more rigorous attribution standards as the race for efficient KV cache compression intensifies.