Refine checkpoint pipeline and card layout

README.md

@@ -27,7 +27,7 @@ tags:
   <strong>Recommended: M · 2.39 GB · 98% of BF16 IFEval</strong>
 </p>
 
-<p align="center"><a href="https://github.com/TheStageAI/edge-lm"><img src="./assets/cta-edge-lm.svg" height="36" alt="Explore edge-lm on GitHub"></a> <a href="https://docs.thestage.ai/"><img src="./assets/cta-docs.svg" height="36" alt="Read TheStageAI documentation"></a> <a href="https://app.thestage.ai/"><img src="./assets/cta-platform.svg" height="36" alt="Open TheStageAI Platform"></a></p>
+<p align="center"><a href="https://github.com/TheStageAI/edge-lm"><img src="./assets/cta-edge-lm.svg" width="104" height="36" alt="Explore edge-lm on GitHub"></a>&nbsp;&nbsp;<a href="https://docs.thestage.ai/"><img src="./assets/cta-docs.svg" width="83" height="36" alt="Read TheStageAI documentation"></a>&nbsp;&nbsp;<a href="https://app.thestage.ai/"><img src="./assets/cta-platform.svg" width="110" height="36" alt="Open TheStageAI Platform"></a></p>
 
 <p align="center"><strong>Qwen 3.5 family:</strong> <a href="https://huggingface.co/TheStageAI/Qwen3.5-0.8B-GGUF">0.8B</a> &nbsp;·&nbsp; <a href="https://huggingface.co/TheStageAI/Qwen3.5-2B-GGUF">2B</a> &nbsp;·&nbsp; <strong><a href="https://huggingface.co/TheStageAI/Qwen3.5-4B-GGUF">4B</a></strong> &nbsp;·&nbsp; <a href="https://huggingface.co/TheStageAI/Qwen3.5-9B-GGUF">9B</a></p>
 
@@ -75,16 +75,21 @@ XS targets minimum footprint for non-thinking chat and instruction following. Fo
 
 </details>
 
-## How these checkpoints were built
+## How we build the checkpoints
 
-The release combines model-specific mixed precision with fixed-precision reference points:
+Every tier starts from [`Qwen/Qwen3.5-4B`](https://huggingface.co/Qwen/Qwen3.5-4B) at pinned revision [`851bf6e8`](https://huggingface.co/Qwen/Qwen3.5-4B/tree/851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0a) and passes through the same five-stage pipeline:
 
-- **XS / S — adaptive.** Per-tensor schedules are selected for explicit size targets.
-- **M / L — fixed precision.** Quantized decoder tensors use Q4_K or Q8_0, respectively.
+| Stage | What happens |
+| --- | --- |
+| **1 · Pin the source** | Lock the upstream weights, tokenizer metadata, and model topology. |
+| **2 · Build the bank** | Produce native Q2_K–Q8_0 candidates for each quantizable tensor group and record their exact byte costs. |
+| **3 · Select and assemble** | Choose a model-specific XS/S assignment under an exact size cap; assemble M/L with fixed Q4_K/Q8_0 assignments. |
+| **4 · Tune scales** | Optimize representable scales and minima against teacher outputs while packed codes, dense weights, and tensor layout remain frozen. |
+| **5 · Validate the release** | Export GGUF, verify size and payload integrity, load with pinned llama.cpp, and run held-out KL before downstream benchmarks. |
 
-Every tier receives scale-only tuning without changing its packed codes or tensor layout. The recommendation is selected independently for each base model from end-to-end evaluation—not from the tier name or nominal bit label.
+**Adaptive schedules for this model:** XS — `RCO anchor`; S — `RCO anchor`. `Uniform` scores module candidates independently; `anchor` uses activations from a concrete model trajectory. The S target matches the exact byte size of a pinned UD-Q2_K_XL checkpoint; its schedule and weights are produced by this pipeline.
 
-**Source:** [`Qwen/Qwen3.5-4B`](https://huggingface.co/Qwen/Qwen3.5-4B) at revision [`851bf6e8`](https://huggingface.co/Qwen/Qwen3.5-4B/tree/851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0a).
+The recommended tier is selected independently for each base model from complete end-to-end evaluation—not from the tier name or nominal bit label.
 
 <details>
 <summary><b>Technical file details</b></summary>

release-manifest.json

@@ -36,7 +36,7 @@
   },
   "display_name": "Qwen3.5 4B",
   "family": "Qwen 3.5",
-  "generated_at": "2026-07-21T00:09:06.790727+00:00",
+  "generated_at": "2026-07-21T10:34:15.339530+00:00",
   "license": "apache-2.0",
   "model_key": "qwen3p5_4b",
   "reasoning_policy": {
@@ -91,6 +91,7 @@
       "product": "XS",
       "reasoning_support": "non_thinking_only",
       "recommended": false,
+      "schedule_method": "RCO anchor",
       "sha256": "a1d49f1d782073748589d0df0893658040231866ae6d69667994bef5cc6ce61f",
       "tensor_count": 426,
       "tensor_type_counts": {
@@ -136,6 +137,7 @@
       "product": "S",
       "reasoning_support": "supported",
       "recommended": false,
+      "schedule_method": "RCO anchor",
       "sha256": "9f020bfc77a9503a641e3e0511de04b4bfba16b9facfa62c3772b4f8e318ba55",
       "tensor_count": 426,
       "tensor_type_counts": {
@@ -183,6 +185,7 @@
       "product": "M",
       "reasoning_support": "supported",
       "recommended": true,
+      "schedule_method": "fixed Q4_K",
       "sha256": "f8e45572b9cc35161d4772b09bccfd383fe0bb03fc6d69b40a9138731302290b",
       "tensor_count": 426,
       "tensor_type_counts": {
@@ -226,6 +229,7 @@
       "product": "L",
       "reasoning_support": "supported",
       "recommended": false,
+      "schedule_method": "fixed Q8_0",
       "sha256": "b81d86c3ccbd1af44474597c0ef97af8e4d71ba40f3b5e7a20b4d6c2f709203c",
       "tensor_count": 426,
       "tensor_type_counts": {