Support transformers 5 and refresh model card

Co-Authored-By: Claude Opus 5.5 <[email protected]>

README.md

@@ -2,7 +2,12 @@
 language: en
 license: apache-2.0
 base_model: google/flan-t5-base
-pipline_tag: text-classficiation
+pipeline_tag: text-classification
+library_name: transformers
+tags:
+  - hallucination-detection
+  - factual-consistency
+  - rag
 
 extra_gated_fields:
   Company: text
@@ -15,11 +20,11 @@ extra_gated_fields:
 
 ## Quickstart: try HHEM-2.1-Open Live Demo
 
-👉 **[Launch Interactive Demo](https://huggingface.co/spaces/vectara/hhem-2.1-open-demo)** - No setup required, runs in your browser
+👉 **[Launch Interactive Demo](https://huggingface.co/spaces/vectara/hhem-2.1-open-demo)** - No setup required, runs in your browser. Paste a source text and an LLM response to get a verdict, a 0-1 consistency score, and per-sentence scores.
 
-<iframe src="https://vectara-hhem-2-1-open-demo.hf.space/" title="Demo for HHEM-2.1-Open"></iframe>
+[![HHEM-2.1-Open hallucination detector demo](https://huggingface.co/spaces/vectara/hhem-2.1-open-demo/resolve/main/assets/thumbnail.png)](https://huggingface.co/spaces/vectara/hhem-2.1-open-demo)
 
-💡 **Quick test**: Try inputting "The capital of France is Berlin" as premise and "The capital of France is Paris" as hypothesis to see HHEM detect this factual but hallucinated case.
+💡 **Quick test**: [Try "The capital of France is Berlin" as the source and "The capital of France is Paris" as the response](https://vectara-hhem-2-1-open-demo.hf.space/?p=The%20capital%20of%20France%20is%20Berlin.&h=The%20capital%20of%20France%20is%20Paris.) to see HHEM detect this factual but hallucinated case.
 
 HHEM-2.1-Open is the latest open source version of [Vectara](https://vectara.com)'s HHEM series models for detecting hallucinations in LLMs. These are particularly useful in the context of building retrieval-augmented-generation (RAG) applications or Agentic workflows, where a set of facts is summarized by an LLM, and HHEM can be used to measure the extent to which this summary is factually consistent with the facts.
 
@@ -31,7 +36,7 @@ For example, given the premise _"The capital of France is Berlin"_, the hypothes
 
 Additionally, hallucination detection is "asymmetric" or is not commutative. For example, the hypothesis _"I visited Iowa"_ is considered hallucinated given the premise _"I visited the United States"_, but the reverse is consistent.
 
-💡 **Using HHEM in production?** We'd love to hear about your use case! Connect with us on [LinkedIn](https://linkedin.com/company/vectara) or [Twitter](https://twitter.com/vectara).
+💡 **Using HHEM in production?** We'd love to hear about your use case! Connect with us on [LinkedIn](https://linkedin.com/company/vectara) or [X](https://x.com/vectara).
 
 ## AI Engineers: Stay Updated on Hallucination Mitigation
 
@@ -51,12 +56,12 @@ Together, these tools provide a comprehensive approach to hallucination mitigati
 
 3. Join our community:
 - 🔗 **LinkedIn**: Follow [@Vectara](https://linkedin.com/company/vectara) for AI safety insights and industry updates
-- 🐦 **X/Twitter**: [@vectara](https://twitter.com/vectara) for real-time updates and research discussions  
+- 🐦 **X**: [@vectara](https://x.com/vectara) for real-time updates and research discussions  
 - ⭐ **GitHub**: [Star our hallucination leaderboard](https://github.com/vectara/hallucination-leaderboard) which tracks LLM hallucination rates across leading LLMs.
 
 ## Using HHEM-2.1-Open
 
-Here we provide several ways to use HHEM-2.1-Open in the `transformers` library.
+Here we provide several ways to use HHEM-2.1-Open in the `transformers` library. Both transformers 4.x and 5.x are supported.
 
 >   You may run into a warning message that "Token indices sequence length is longer than the specified maximum sequence length". Please ignore it which is inherited from the foundation, T5-base. 
 
@@ -128,7 +133,7 @@ Of course, with `pipeline`, you can also get the most likely label, or the label
 
 ## HHEM-2.3 and the LLM Hallucination Leaderboard
 
-**See how LLMs compare**: HHEM-2.3, our latest commercial hallucination detection model, powers our [live leaderboard](https://huggingface.co/spaces/vectara/leaderboard) that continuously benchmarks leading LLMs for hallucination rates. Watch comparisons of the latest models including GPT-5, Claude 4, Grok-4, Gemini 2.5, Llama4, Mistrak, DeepSeek, and many others.
+**See how LLMs compare**: HHEM-2.3, our latest commercial hallucination detection model, powers our [live leaderboard](https://huggingface.co/spaces/vectara/leaderboard) that continuously benchmarks leading LLMs for hallucination rates. Compare the latest models including GPT-6, GPT-5.x, Claude Opus 4.7 and Sonnet 4.6, Gemini 3.1, Grok 4.1, Gemma 4, Llama 4, Mistral, DeepSeek, Qwen 3, and many others.
 
 **HHEM-2.3 advantages over the open source version**:
 - Enhanced accuracy and performance
@@ -185,9 +190,9 @@ Another advantage of HHEM-2.1-Open is its efficiency. HHEM-2.1-Open can be run o
 
 ## Hallucination detection with Vectara
 
-Vectara provides a Trusted Generative AI platform. The platform allows organizations to rapidly create an AI agent experience which is grounded in the data, documents, and knowledge that they have. Vectara solves critical problems required for enterprise adoption of RAG and Agentic AI applications, namely: reduces hallucination, provides explainability / provenance, enforces access control, allows for real-time updatability of the knowledge, and mitigates intellectual property / bias concerns from large language models.
+Vectara is an enterprise agent platform that powers intelligent, managed AI agents grounded in an organization's own data, documents, and knowledge, including complex multimodal data. Vectara solves critical problems required for enterprise adoption of RAG and Agentic AI applications, namely: reduces hallucination, provides explainability / provenance, enforces access control, allows for real-time updatability of the knowledge, and mitigates intellectual property / bias concerns from large language models.
 
-HHEM-2.3 is fully integrated into Vectara and is automtically returned with every query API call.
+HHEM-2.3 is fully integrated into Vectara and is automatically returned with every query API call.
 
 To start benefiting from HHEM-2.3, you can [sign up](https://console.vectara.com/signup/?utm_source=huggingface&utm_medium=space&utm_term=hhem-model&utm_content=console&utm_campaign=) for a Vectara account, and you will get the HHEM-2.3 score returned with every query automatically.
 

modeling_hhem_v2.py

@@ -30,6 +30,7 @@ class HHEMv2ForSequenceClassification(PreTrainedModel):
         )
         self.prompt = config.prompt
         self.tokenzier = AutoTokenizer.from_pretrained(config.foundation)
+        self.post_init() # required by transformers>=5 (sets all_tied_weights_keys)
 
     def populate(self, model: AutoModel): 
         """Initiate the model with the pretrained model