arXiv:2606. 09012v1 Announce Type: cross Abstract: Post-training quantization (PTQ) converts a trained full-precision model into low-bit weights without task-level retraining, while quantization-aware training (QAT) incorporates quantization into the training loop.
As large language model inference shifts toward lower precision, post-training quantization (PTQ) becomes increasingly brittle, making quantization-aware training (QAT) essential for preserving model...
arXiv:2608. 13966v1 Announce Type: new Abstract: As large language model inference shifts toward lower precision, post-training quantization (PTQ) becomes increasingly brittle, making quantization-aware training (QAT) essential for preserving model quality.
Post‑training quantization compresses large language models by storing weights at reduced precision, introducing errors into hidden states that could accumulate with depth. However, pretrained models accumulate far less hidden‑state error than randomly initialized ones, largely preserving downstream performance. The study identifies two key mechanisms: (1) each layer’s new error tends to oppose inherited error, partially canceling it, and (2) the LM‑head geometry preserves high‑rank token scores, mitigating output changes.
arXiv:2608. 11045v1 Announce Type: new Abstract: ReRound (Reconstructive Rounding) is a post-training quantization method that addresses the midpoint ambiguity inherent in standard round-to-nearest (RTN) schemes when quantizing weights near the centers of quantization intervals.
arXiv:2510. 18784v3 Announce Type: replace Abstract: Despite significant work on low-bit quantization-aware training (QAT), there is still an accuracy gap between such techniques and native training.
The paper examines weight-only post‑training quantization (PTQ) and finds that minimizing reconstruction loss does not always improve model performance, sometimes even harming it on calibration data. It shows that weights with low reconstruction loss on one distribution can perform worse when activation distributions shift. To address this, the authors introduce Distributionally Robust Quantization (DRQ), a refinement step that minimizes worst‑case reconstruction loss over a constrained set of activation distributions while keeping the quantization grid unchanged. Experiments demonstrate that DRQ improves models quantized by several PTQ methods across dense and mixture‑of‑experts large language models, offering better downstream performance without extra inference cost.
The paper investigates how the choice of reconstruction loss, specifically mean squared error (MSE), affects optimization in sequential post‑training quantization (PTQ) of large language models (LLMs). It identifies an Optimization Imbalance where MSE causes reconstruction loss magnitudes—and thus gradient magnitudes—to vary dramatically across quantization stages, leading to uneven parameter updates. The authors propose that root mean squared error (RMSE) variants, which implicitly normalize gradients, can decouple optimization strength from loss scale and outperform MSE as a drop‑in replacement.
REAL-Q introduces a new post‑training quantization approach for large language models that replaces the traditional single closed‑form second‑order solver with a fine‑grained, dynamic block‑wise gradient descent applied after every 128‑column block. By aligning the surrogate loss with the end‑to‑end objective and using a sliding window for smooth cross‑layer transitions, REAL‑Q mitigates error propagation and information misalignment. Experiments on LLaMA‑3.1 and Qwen3 show up to ~49% reduction in end‑to‑end KL divergence compared to state‑of‑the‑art methods.
arXiv:2607. 07964v1 Announce Type: new Abstract: Post-training quantization (PTQ) is a widely adopted technique for compressing large language models (LLMs) without retraining.
Q-PACE is a dynamic precision allocation method for quantization-aware training that uses a second-order sensitivity model to predict loss increases from quantization noise. By periodically recalculating per-layer curvature coefficients and reassigning precision, it improves mixed-precision training for large language models up to 4 B parameters, achieving comparable loss with lower memory usage. The approach shows that quantization sensitivity is predictable by depth and layer type, enabling infrequent, inexpensive recalibration.
arXiv:2609.30820v1 Announce Type: cross Abstract: Looped transformers reuse weights across recurrence steps, making low-bit quantization especially attractive. We identify two distinct failure modes...
arXiv:2601. 21626v2 Announce Type: replace-cross Abstract: Post Training Quantization (PTQ), a mainstream model compression technique, often leads to the paradoxical 'low error, high loss' phenomenon because it focuses solely on minimizing quantization error.