Add a known runtime issue: the stop token with a zero embedding from the second turn

README.md

@@ -141,6 +141,10 @@ For the Chinese example clip `raw` is `language Chinese<asr_text>开放时间：
 
 Sample app (Android): [`android/`](https://huggingface.co/mlboydaisuke/Confucius4-R2T2-LiteRT/tree/main/android) — one app for this file and two sibling bundles (Qwen3-ASR-1.7B, Fun-ASR-Nano-2512, Confucius4-R2T2): a model picker, hold-to-talk up to 30 s and three FLEURS clips, with a device check. Galaxy S26, language model and audio encoder on the CPU (4 threads), the 6.4 s English clip in 2.59 s, the 9.7 s Mandarin clip in 3.48 s and the 11.0 s Japanese clip in 3.54 s; a 9 s microphone recording in 3.17 s with no character error, measured by the app's device check on 2026-10-10 (CPU at its full clock).
 
+## Known runtime issues
+
+From the second turn of a conversation on, LiteRT-LM puts the previous answer's stop token into the KV cache with a zero embedding, because this file holds the embedding table in its own section and the decoder takes embeddings, so later answers in a conversation can differ from the answer the model gives to the same messages in one prompt (read in the source of LiteRT-LM 0.16.0 to 0.18.1; measured with 0.18.1 on WeVisOne-VL-2B and Qwen2.5-Coder-1.5B-Instruct, not on this file). One question per conversation avoids it, as does a new conversation that starts with the earlier messages and the new question.
+
 ## License
 
 This file is a quantized Derivative Work of Confucius4-R2T2 and is distributed under the **NetEase Youdao Model Use License Agreement**. The agreement is included in this repository in English (`MODEL_LICENSE`) and Chinese (`MODEL_LICENSE_zh`); clause 9 makes the Chinese version prevail.