Rebrand to While AI

README.md

@@ -2,7 +2,7 @@
 license: apache-2.0
 base_model: Qwen/Qwen2.5-0.5B-Instruct
 datasets:
-  - zero-proof-ai/ecommerce-intent
+  - while-ai/ecommerce-intent
 library_name: peft
 pipeline_tag: text-classification
 tags:
@@ -15,9 +15,9 @@ language:
   - en
 ---
 
-# Model Card for zeroproof-ecommerce-0.5b
+# Model Card for ecommerce-0.5b
 
-**E-commerce intent detection in a 0.5B model, the smallest member of the Zero Proof family.** Given a live customer conversation, `zeroproof-ecommerce-0.5b` returns a single JSON object with the customer's payment intent and its structured details, at the lowest cost and latency in the family.
+**E-commerce intent detection in a 0.5B model, the smallest member of the While AI family.** Given a live customer conversation, `ecommerce-0.5b` returns a single JSON object with the customer's payment intent and its structured details, at the lowest cost and latency in the family.
 
 On a hard, held-out benchmark it reaches **70.3 macro-averaged intent-type accuracy, roughly 91% of our 1B**, from a base model that scores 10.7.
 
@@ -25,7 +25,7 @@ On a hard, held-out benchmark it reaches **70.3 macro-averaged intent-type accur
 
 ### Model Description
 
-- **Developed by:** Zero Proof Labs
+- **Developed by:** While AI
 - **Model type:** E-commerce payment-intent classifier; structured JSON output over seven intent types
 - **Language:** English
 - **License:** Apache-2.0 (inherited from the base model)
@@ -33,8 +33,8 @@ On a hard, held-out benchmark it reaches **70.3 macro-averaged intent-type accur
 
 ### Model Sources
 
-- **Repository:** https://huggingface.co/zero-proof-ai/zeroproof-ecommerce-0.5b
-- **Training data:** https://huggingface.co/datasets/zero-proof-ai/ecommerce-intent
+- **Repository:** https://huggingface.co/while-ai/ecommerce-0.5b
+- **Training data:** https://huggingface.co/datasets/while-ai/ecommerce-intent
 
 ## Uses
 
@@ -60,8 +60,8 @@ Pair the model with a verification layer before any payment executes, and re-ben
 from peft import AutoPeftModelForCausalLM
 from transformers import AutoTokenizer
 
-model = AutoPeftModelForCausalLM.from_pretrained("zero-proof-ai/zeroproof-ecommerce-0.5b")
-tok   = AutoTokenizer.from_pretrained("zero-proof-ai/zeroproof-ecommerce-0.5b")
+model = AutoPeftModelForCausalLM.from_pretrained("while-ai/ecommerce-0.5b")
+tok   = AutoTokenizer.from_pretrained("while-ai/ecommerce-0.5b")
 ```
 
 Use greedy decoding (`temperature=0`). The system prompt and per-intent detail schemas ship in the repo.
@@ -70,7 +70,7 @@ Use greedy decoding (`temperature=0`). The system prompt and per-intent detail s
 
 ### Training Data
 
-The differentiator is the data. Zero Proof Labs builds e-commerce intent training data as **randomized conversational simulations**: role-played customers with independently sampled personas, situations, and behaviors including adversarial actors, simulated turn by turn between two independently drawn models. Generation is label-blind, labels are assigned in a separate pass, and every candidate passes a structural data gate with zero train/eval leakage. This 0.5B was trained on the same class-rebalanced split as the 1B so it routes across all intent types rather than defaulting to the majority.
+The differentiator is the data. While AI builds e-commerce intent training data as **randomized conversational simulations**: role-played customers with independently sampled personas, situations, and behaviors including adversarial actors, simulated turn by turn between two independently drawn models. Generation is label-blind, labels are assigned in a separate pass, and every candidate passes a structural data gate with zero train/eval leakage. This 0.5B was trained on the same class-rebalanced split as the 1B so it routes across all intent types rather than defaulting to the majority.
 
 ### Training Procedure
 
@@ -88,7 +88,7 @@ Held-out eval of 1,977 conversations, zero train/eval leakage, macro-averaged, s
 
 ### Results
 
-![zeroproof-ecommerce-0.5b vs its base and frontier models on e-commerce intent](0.5b-v2.png)
+![ecommerce-0.5b vs its base and frontier models on e-commerce intent](0.5b-v2.png)
 
 - Fine-tuning takes intent-type accuracy from **10.7 to 70.3** (macro), more than 6x the base, at ~91% of our 1B and a fraction of frontier size.
 - On structured detail extraction the 0.5B **beats every frontier panel member**: 58.5 against Sonnet 5's 55.7, Opus 4.8's 53.2, and GPT-5's 41.4. The 1B leads the family at 64.0.
@@ -102,4 +102,4 @@ A LoRA adapter over `Qwen2.5-0.5B-Instruct`, trained response-only to emit one s
 
 ## Model Card Contact
 
-Zero Proof Labs, https://huggingface.co/zero-proof-ai
+While AI, https://huggingface.co/while-ai