CarlosAGDev
/

ltv-lora-qa

@@ -1,199 +1,115 @@
 ---
-library_name: transformers
-tags: []
 ---
-# Model Card for Model ID
-<!-- Provide a quick summary of what the model is/does. -->
-## Model Details
-### Model Description
-<!-- Provide a longer summary of what this model is. -->
-This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
-- **Developed by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Model type:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** [More Information Needed]
-- **Finetuned from model [optional]:** [More Information Needed]
-### Model Sources [optional]
-<!-- Provide the basic links for the model. -->
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
-## Uses
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
-### Direct Use
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
-[More Information Needed]
-### Downstream Use [optional]
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
-[More Information Needed]
-### Out-of-Scope Use
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
-[More Information Needed]
-## Bias, Risks, and Limitations
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
-[More Information Needed]
-### Recommendations
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
-## How to Get Started with the Model
-Use the code below to get started with the model.
-[More Information Needed]
-## Training Details
-### Training Data
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
-[More Information Needed]
-### Training Procedure
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
-#### Preprocessing [optional]
-[More Information Needed]
-#### Training Hyperparameters
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
-#### Speeds, Sizes, Times [optional]
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
-[More Information Needed]
-## Evaluation
-<!-- This section describes the evaluation protocols and provides the results. -->
-### Testing Data, Factors & Metrics
-#### Testing Data
-<!-- This should link to a Dataset Card if possible. -->
-[More Information Needed]
-#### Factors
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
-[More Information Needed]
-#### Metrics
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
-[More Information Needed]
-### Results
-[More Information Needed]
-#### Summary
-## Model Examination [optional]
-<!-- Relevant interpretability work for the model goes here -->
-[More Information Needed]
-## Environmental Impact
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
-## Technical Specifications [optional]
-### Model Architecture and Objective
-[More Information Needed]
-### Compute Infrastructure
-[More Information Needed]
-#### Hardware
-[More Information Needed]
-#### Software
-[More Information Needed]
-## Citation [optional]
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
-**BibTeX:**
-[More Information Needed]
-**APA:**
-[More Information Needed]
-## Glossary [optional]
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
-[More Information Needed]
-## More Information [optional]
-[More Information Needed]
-## Model Card Authors [optional]
-[More Information Needed]
-## Model Card Contact
-[More Information Needed]

 ---
+base_model: google/gemma-4-E2B-it
+library_name: peft
+license: apache-2.0
+tags:
+- lora
+- qlora
+- gemma
+- ltv
+- question-generation
 ---
+# CarlosAGDev/ltv-lora-qa
+LoRA-QA (v0.1.0) del LTV Framework. Genera 2-4 sub-preguntas atomicas y
+auto-contenidas para verificar una afirmacion check-worthy clasificada. Tercer
+paso del pipeline de Triage, despues de LoRA-CW y LoRA-CLF.
+## Formato de salida
+```json
+{
+  "questions": [
+    {"question": "...", "answer_type": "Boolean"},
+    {"question": "...", "answer_type": "Extractive"}
+  ]
+}
+```
+`answer_type` puede ser: `Boolean`, `Extractive`, o `Abstractive`.
+## Detalles del Entrenamiento
+Entrenado sobre anotaciones sinteticas generadas por `gemini-3.1-flash-lite`.
+Pool balanceado: Event/Property Claim capeado a 500 (de 973 disponibles), split
+estratificado por claim_type (15% eval, 85% train).
+| Claim type entrenado | Claims (pool) |
+| :--- | :---: |
+| Event/Property Claim | 500 (cap desde 973) |
+| Numerical Claim | 207 |
+| Causal Claim | 12 |
+| Position Statement | 8 |
+| Quote Verification | 0 (ausente en v0.1) |
+| **Total** | **727** |
+### Hiperparametros
+- **Modelo Base:** `google/gemma-4-E2B-it`
+- **Max Sequence Length:** `512`
+- **Epochs:** `1`
+- **Batch Size (Per Device):** `4`
+- **Gradient Accumulation Steps:** `4`
+- **Learning Rate:** `0.0002`
+- **Optimizer:** `paged_adamw_8bit`
+## Resultados (v0.1.0)
+Evaluado sobre 110 muestras (15% held-out, estratificado por claim_type).
+Tiempo de evaluacion: 32m 6s (~17.5 s/ejemplo).
+### Validez del esquema de salida
+| Metrica | Valor |
+| :--- | :--- |
+| **JSON valido + schema OK** | **110/110 (100%)** |
+| **Fallos de parseo** | **0** |
+### Validez por claim_type
+| Claim type | Validos / Total | % |
+| :--- | :---: | :---: |
+| Event/Property Claim | 76/76 | 100% |
+| Numerical Claim | 31/31 | 100% |
+| Causal Claim | 2/2 | 100% |
+| Position Statement | 1/1 | 100% |
+### Distribucion de n_questions (Referencia vs Generado)
+| n preguntas | Referencia | Generado |
+| :---: | :---: | :---: |
+| 2 | 14 | 2 |
+| 3 | 92 | **107** |
+| 4 | 4 | 1 |
+### Distribucion de answer_type (Referencia vs Generado)
+| answer_type | Ref count | Ref % | Gen count | Gen % |
+| :--- | :---: | :---: | :---: | :---: |
+| Boolean | 130 | 40.6% | 127 | 38.6% |
+| Extractive | 146 | 45.6% | 171 | **52.0%** |
+| Abstractive | 44 | 13.8% | 31 | 9.4% |
+### Notas de comportamiento
+- **Generacion perfecta en v0.1:** el modelo produce JSON valido y schema-correcto en
+  el 100% de los casos evaluados, para los 4 tipos de claim disponibles.
+- **Preferencia por 3 preguntas:** el modelo genera 3 sub-preguntas en el 97% de los
+  casos (referencia: 84%), colapsando los extremos (2 y 4 preguntas se usan mucho menos).
+  Para la mayoria de afirmaciones esto es correcto, pero puede infragenerar preguntas
+  para claims complejos que merecen 4.
+- **Sesgo Extractive / falta de Abstractive:** genera Extractive un 6% mas de lo esperado
+  (52% vs 45.6%) y Abstractive un 4% menos (9.4% vs 13.8%). El modelo favorece preguntas
+  de hecho concreto sobre preguntas de sintesis o explicacion causal.
+### Limitaciones y mejoras para v0.2.0
+1. **Quote Verification ausente:** 0 ejemplos en el batch sintetico actual. Al completar
+   las 7,440 anotaciones aparecera este tipo. Hasta entonces el modelo no ha aprendido a
+   generar preguntas para citas textuales.
+2. **Causal y Position muy escasos (12 y 8 claims):** los resultados 100% para estos tipos
+   son prometedores pero no estadisticamente robustos con 2 y 1 ejemplos de eval.
+3. **Distribucion de n_questions sesgada hacia 3:** agregar ejemplos con 2 y 4 preguntas
+   equilibrara la distribucion generada.
+4. **Reducir sesgo Extractive:** el batch sintetico completo dara mas ejemplos con
+   Abstractive (tipicamente en Causal y Position Statement).