Update paper metadata: new title, Spellbrush-only authors, arXiv-preprint citation

Match the arXiv submission: title changed to "When Does Predictor-Based RL
Align with Human Perception?", author line added (Park, Li — Spellbrush),
EMNLP-2026 inproceedings citation replaced with an arXiv @misc entry pending
the assigned identifier.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MUq8H7Bue5vjhcuKWSgZNs

Author: nonmetal

README.md

@@ -2,8 +2,9 @@
 
 **Constrained perceptual GRPO for codec-based speech language models.**
 
-Companion code for *"When Do Perceptual Rewards Align Speech Language Models?
-Constrained GRPO for Subjective Style Control"* (EMNLP 2026).
+Companion code for *"When Does Predictor-Based RL Align with Human Perception?
+A Study of Subjective Rewards in Codec-Based Speech Language Models"*
+(Joonyong Park and Jerry Li, Spellbrush — arXiv preprint, 2026).
 
 🔊 **[Audio demo →](https://sizigi.github.io/animeGRPO/)**
 
@@ -183,14 +184,17 @@ zone gate — the ablation in the paper is exactly "swap the axis, keep the gate
 ## Citation
 
 ```bibtex
-@inproceedings{animegrpo2026,
-  title     = {When Do Perceptual Rewards Align Speech Language Models?
-               Constrained {GRPO} for Subjective Style Control},
-  booktitle = {Proceedings of EMNLP 2026},
-  year      = {2026}
+@misc{park2026predictorrl,
+  title  = {When Does Predictor-Based {RL} Align with Human Perception?
+            A Study of Subjective Rewards in Codec-Based Speech Language Models},
+  author = {Park, Joonyong and Li, Jerry},
+  year   = {2026},
+  note   = {arXiv preprint}
 }
 ```
 
+The arXiv identifier will be added to the entry once assigned.
+
 ## License
 
 Code is Apache-2.0 ([LICENSE](LICENSE)), matching the verl fork it extends.

docs/index.html

@@ -173,11 +173,11 @@
 <header>
 <div class="wrap">
   <p class="eyebrow">animeGRPO</p>
-  <h1>When Do Perceptual Rewards Align Speech Language Models?
-    <span class="sub">Constrained GRPO for subjective style control in codec speech LMs</span>
+  <h1>When Does Predictor-Based RL Align with Human Perception?
+    <span class="sub">A study of subjective rewards in codec-based speech language models</span>
   </h1>
-  <!-- Author list intentionally left out until the paper is public. -->
-  <p class="venue">EMNLP 2026 &nbsp;·&nbsp; LLaSA-1B-Multilingual &nbsp;·&nbsp; Japanese &amp; English</p>
+  <p class="venue">Joonyong Park &nbsp;·&nbsp; Jerry Li &nbsp;&mdash;&nbsp; Spellbrush</p>
+  <p class="venue" style="margin-top:6px">arXiv preprint &nbsp;·&nbsp; LLaSA-1B-Multilingual &nbsp;·&nbsp; Japanese &amp; English</p>
   <div class="links">
     <a class="btn" href="https://github.com/sizigi/animeGRPO">
       <svg viewBox="0 0 16 16"><path d="M8 0C3.58 0 0 3.58 0 8c0 3.54 2.29 6.53 5.47 7.59.4.07.55-.17.55-.38 0-.19-.01-.82-.01-1.49-2.01.37-2.53-.49-2.69-.94-.09-.23-.48-.94-.82-1.13-.28-.15-.68-.52-.01-.53.63-.01 1.08.58 1.23.82.72 1.21 1.87.87 2.33.66.07-.52.28-.87.51-1.07-1.78-.2-3.64-.89-3.64-3.95 0-.87.31-1.59.82-2.15-.08-.2-.36-1.02.08-2.12 0 0 .67-.21 2.2.82a7.4 7.4 0 0 1 2-.27c.68 0 1.36.09 2 .27 1.53-1.04 2.2-.82 2.2-.82.44 1.1.16 1.92.08 2.12.51.56.82 1.27.82 2.15 0 3.07-1.87 3.75-3.65 3.95.29.25.54.73.54 1.48 0 1.07-.01 1.93-.01 2.2 0 .21.15.46.55.38A8.01 8.01 0 0 0 16 8c0-4.42-3.58-8-8-8z"/></svg>
@@ -338,11 +338,12 @@ <h2>Listen</h2>
 
 <footer class="wrap">
   <h2 style="margin-bottom:16px">Citation</h2>
-  <pre class="cite">@inproceedings{animegrpo2026,
-  title     = {When Do Perceptual Rewards Align Speech Language Models?
-               Constrained {GRPO} for Subjective Style Control},
-  booktitle = {Proceedings of EMNLP 2026},
-  year      = {2026}
+  <pre class="cite">@misc{park2026predictorrl,
+  title  = {When Does Predictor-Based {RL} Align with Human Perception?
+            A Study of Subjective Rewards in Codec-Based Speech Language Models},
+  author = {Park, Joonyong and Li, Jerry},
+  year   = {2026},
+  note   = {arXiv preprint}
 }</pre>
   <p>Code is Apache-2.0. Audio and score tables are released for research
   inspection and replication of the paper's results.</p>