Qwen3.8-27B-GSQ-RCO-BuffedMod-GGUF

Qwen3.8-27B-GSQ-RCO — BuffedMod GGUF

This repository contains modified versions of the Qwen3.8-27B-GSQ-RCO-IQ3_S GGUF model.

Unlike traditional re-quantization workflows which require quantizing the entire model from unquantized FP16/BF16 weights, these files were created by directly modifying specific tensor weights within the binary GGUF file.

By replacing critical layer tensors (specifically output.weight) with higher-precision quantizations, we achieve better Perplexity (PPL), lower KL Divergence (KLD), and increased top-token agreement while maintaining a compact footprint.

---

⚡ What Changed in BuffedMod?

In lower-bit quantizations like IQ3_S, the output.weight tensor (LM Head) often serves as a major bottleneck for final text generation precision. Upgrading this tensor improves top-token probability matching without adding massive overhead to the overall file size.

📦 Available Variants

1. Qwen3.8-27B-GSQ-RCO-BuffedMod-IQ3_S_XL-mtp-Q6K.gguf

   • Modifications: Replaced output.weight from Q4_K to Q6_K.
   • Target: Maximum quality retention for IQ3_S baseline.

2. Qwen3.8-27B-GSQ-RCO-BuffedMod-IQ3_S_XL-mtp-IQ4XS.gguf

   • Modifications: Replaced output.weight from Q4_K to Q6_K AND down-quantized MTP (Multi-Token Prediction) tensors from Q6_K to IQ4_XS.
   • Target: Optimized size-to-performance balance by reclaiming memory from non-essential auxiliary tensors.

---

📊 Benchmark & Evaluation Results

All variants were evaluated using llama-perplexity with original BF16 base model (q38f16baseline.kld) to measure Perplexity (PPL) degradation, KL Divergence (KLD), and Token Probability shifts.

llama-perplexity.exe -f /wikitext-2-raw/wiki.test.raw \
    --kl-divergence \
    --kl-divergence-base q38f16baseline.kld \
    -ngl 99 \
    -m <model_variant>.gguf

📈 Comparison Summary

---

📐 Model Variant Comparison

The table below highlights the trade-off between file size efficiency and accuracy gains across the original and BuffedMod variants:

📌 Key Highlights:

• BuffedMod (IQ4XS MTP) offers the optimal sweet spot: it gains the full accuracy boost (+0.88% top-token match, -27.2% PPL error vs base) with only +200 MB size overhead compared to the original file.
• BuffedMod (Q6_K MTP) retains maximum fidelity on the auxiliary MTP tensors for an extra +100 MB, maintaining identical core perplexity and divergence performance.

📋 Full Metrics Breakdown

🔴 Baseline Original Metrics (IQ3_S with Q4_K output.weight)

====== Perplexity statistics ======
Mean PPL(Q)                   :   7.062697 ±   0.045484
Mean PPL(base)                :   6.950493 ±   0.044934
Cor(ln(PPL(Q)), ln(PPL(base))):  98.65%
Mean ln(PPL(Q)/PPL(base))     :   0.016014 ±   0.001059
Mean PPL(Q)/PPL(base)         :   1.016143 ±   0.001076
Mean PPL(Q)-PPL(base)         :   0.112205 ±   0.007438

====== KL divergence statistics ======
Mean    KLD:   0.055475 ±   0.000453
Maximum KLD:  15.811345
99.9%   KLD:   2.067696
99.0%   KLD:   0.593792
95.0%   KLD:   0.188752
90.0%   KLD:   0.110145
Median  KLD:   0.022928

====== Token probability statistics ======
Mean    Δp: -0.513 ± 0.017 %
RMS Δp    :  6.537 ± 0.051 %
Same top p: 89.657 ± 0.079 %

🟢 BuffedMod Metrics (IQ3_S with Q6_K output.weight)

====== Perplexity statistics ======
Mean PPL(Q)                   :   7.032142 ±   0.045271
Mean PPL(base)                :   6.950493 ±   0.044934
Cor(ln(PPL(Q)), ln(PPL(base))):  98.72%
Mean ln(PPL(Q)/PPL(base))     :   0.011679 ±   0.001034
Mean PPL(Q)/PPL(base)         :   1.011747 ±   0.001046
Mean PPL(Q)-PPL(base)         :   0.081650 ±   0.007236

====== KL divergence statistics ======
Mean    KLD:   0.051738 ±   0.000450
Maximum KLD:  15.754954
99.9%   KLD:   2.056773
99.0%   KLD:   0.586696
95.0%   KLD:   0.181908
90.0%   KLD:   0.103020
Median  KLD:   0.018899

====== Token probability statistics ======
Mean    Δp: -0.477 ± 0.016 %
RMS Δp    :  6.285 ± 0.052 %
Same top p: 90.535 ± 0.076 %

📜 Credits & Acknowledgments

• Original Model: ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF
• Modification: Tweak & Direct Tensor Modification by tooltd