Model card: rename Trida2.0-4B to Zoro (#1)

- Model card: rename Trida2.0-4B to Zoro (d51eec766e9040014207e7f6980b4bcb44ca63cb)

README.md

@@ -8,11 +8,11 @@ tags:
   - qwen3.5
 ---
 
-# trillionlabs/Trida2.0-4B
+# trillionlabs/Zoro
 
 *English | [한국어](./README_ko.md)*
 
-**Trida2.0-4B** is a two-stream **block-diffusion** language model (research checkpoint). It is produced
+**Zoro** is a two-stream **block-diffusion** language model (research checkpoint). It is produced
 by continued SFT with the **full-mask recipe** — 100% of each block masked (single view, no
 complementary mask) at block size **bd=8**, fresh optimizer, LR 5e-5, ~2k steps — starting from the
 base SFT checkpoint (v6, step_18000).
@@ -36,7 +36,7 @@ untrained). Runs on the HybridDiffusion SGLang fork. `mask_id = 248077`, vocab 2
 
 Concurrency 1, threshold 0.9, reasoning on. *A vanilla Qwen3.5-4B baseline is being added.*
 
-| benchmark | Trida2.0 · self-spec | Trida2.0 · AR |
+| benchmark | Zoro · self-spec | Zoro · AR |
 |---|---|---|
 | GSM8K | **84.2** | **89.3** |
 | MMLU-Pro | 63.0 | — |
@@ -49,15 +49,15 @@ Concurrency 1, threshold 0.9, reasoning on. *A vanilla Qwen3.5-4B baseline is be
 ## Serving speed & parallel commit (concurrency 1)
 
 Block-diffusion self-spec commits **multiple tokens per forward pass** at verify-guaranteed AR
-quality. Trida2.0-4B's full-mask training makes it a better drafter, so it accepts more per step and
+quality. Zoro's full-mask training makes it a better drafter, so it accepts more per step and
 runs faster than the base.
 
 | model · decoding | tok/s (per-request) | tokens / forward |
 |---|---|---|
-| **Trida2.0 · self-spec** | **243** | **2.59** |
-| Trida2.0 · AR | 216 | 1.00 |
+| **Zoro · self-spec** | **243** | **2.59** |
+| Zoro · AR | 216 | 1.00 |
 
-Per-benchmark commit rate (Trida2.0 self-spec, tokens/forward): **GSM8K 3.10, BFCL 2.33, MMLU-Pro 2.33**.
+Per-benchmark commit rate (Zoro self-spec, tokens/forward): **GSM8K 3.10, BFCL 2.33, MMLU-Pro 2.33**.
 
 
 ## License