Research map / Architectures and efficiency / Quantization
BBQ: Boosting Quantization Entropy with Bell Box Quantization
Ningfeng Yang, Tor M. Aamodt
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Hadamard Transform (ablation) | BBQ · 2-bit LLaMA-95M, 3B tokens | perplexity: 35.79 -> 31.34 | Improved |
| add RMS normalization (ablation) | BBQ · 2-bit LLaMA-95M, 3B tokens | perplexity: 35.93 -> 31.34 | Improved |
| substitute clip with probability integral transform without gamma init (ablation) | QuEST · 2-bit LLaMA-95M, 3B tokens | perplexity: 35.58 -> 138.3 | Hurt |
| add PIT together with gamma initialization instead of clip (ablation) | QuEST · 2-bit LLaMA-95M, 3B tokens | perplexity: 35.58 -> 31.46 | Improved |
| make gamma learnable (ablation) | BBQ · 2-bit LLaMA-95M, 3B tokens | perplexity: 31.46 -> 31.34 | Improved |
| use BBQ instead of QuEST at 1-bit | QuEST · C4 LLaMA-95M W1A1 | perplexity: 67.78 -> 49.22 | Improved |
| use BBQ instead of QuEST at 2-bit | QuEST · C4 LLaMA-200M W2A2 | perplexity: 25.46 -> 23.08 | Improved |
| use BBQ instead of QuEST at 4-bit | QuEST · C4 LLaMA-125M W4A4 | perplexity: 22.98 -> 22.15 | Improved |
| use BBQ instead of LSQ at 2-bit | LSQ · C4 LLaMA-95M W2A2 | perplexity: 36.58 -> 31.34 | Improved |
| use BBQ instead of QuEST at 2-bit | QuEST · C4 GPT 95M W2A2 | perplexity: 36.7 -> 34.36 | Improved |
| use BBQ instead of N2UQ at 2-bit | N2UQ · C4 GPT 95M W2A2 | perplexity: 35.59 -> 34.36 | Improved |
| use BBQ instead of NF4 for 4-bit inference | NF4 · LLaMA prefill latency | latency: 48% speedup over NF4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- QuEST — QUEST: Query-Aware Sparsity for Efficient Long-Context LLM Inference (ICML 2024) · better on 13 of 13 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Compute-Optimal Quantization-Aware Training (ICLR 2026 · similarity 0.88)
- 1-Bit Wonder: Improving QAT Performance in the Low-Bit Regime through K-Means Quantization (ICML 2026 · similarity 0.88)
- ParetoQ: Improving Scaling Laws in Extremely Low-bit LLM Quantization (NeurIPS 2025 · similarity 0.87)
- A2Q+: Improving Accumulator-Aware Weight Quantization (ICML 2024 · similarity 0.87)
- QBB: Quantization with Binary Bases for LLMs (NeurIPS 2024 · similarity 0.86)
- Q-Palette: Fractional-Bit Quantizers Toward Optimal Bit Allocation for Efficient LLM Deployment (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
