README.md

---

language:
- en
license: apache-2.0
library_name: sentence-transformers
tags:
- sentence-transformers
- sentence-similarity
- feature-extraction
- dense-retrieval
- information-retrieval
- job-skill-matching
- esco
- talentclef
- xlm-roberta
base_model: jjzha/esco-xlm-roberta-large
pipeline_tag: sentence-similarity
model-index:
- name: skillscout-large
  results:
  - task:
      type: information-retrieval
      name: Information Retrieval
    dataset:
      name: TalentCLEF 2026 Task B Validation
      type: talentclef-2026-taskb-validation
    metrics:
    - type: cosine_ndcg_at_10
      value: 0.4830
      name: nDCG@10
    - type: cosine_map_at_100
      value: 0.1825
      name: MAP@100
    - type: cosine_mrr_at_10
      value: 0.6657
      name: MRR@10
    - type: cosine_accuracy_at_10
      value: 0.9474
      name: Accuracy@10
---


# SkillScout Large - Job-to-Skill Dense Retriever

**SkillScout Large** is a dense bi-encoder for retrieving relevant skills from a job title.
Given a job title (e.g., *"Data Scientist"*), it produces a 1024-dimensional embedding and
retrieves the most semantically relevant skills from the [ESCO](https://esco.ec.europa.eu/)
skill gazetteer (9,052 skills) via cosine similarity.

This is **Stage 1** of the TalentGuide two-stage job-skill matching pipeline, trained for
[TalentCLEF 2026 Task B](https://talentclef.github.io/).

> **Best pipeline result (TalentCLEF 2026 validation set):**
> nDCG@10 graded = **0.6896** | nDCG@10 binary = **0.7330**
> when combined with a fine-tuned cross-encoder at blend alpha=0.7.
> Bi-encoder alone: nDCG@10 graded = **0.3621** | MAP = **0.4545**

---

## Model Summary

| Property | Value |
|---|---|
| Base model | [jjzha/esco-xlm-roberta-large](https://huggingface.co/jjzha/esco-xlm-roberta-large) |
| Architecture | XLM-RoBERTa-large + mean pooling |
| Embedding dimension | 1024 |
| Max sequence length | 64 tokens |
| Training loss | Multiple Negatives Ranking (MNR) |
| Training pairs | 93,720 (ESCO job-skill pairs, essential + optional) |
| Epochs | 3 |
| Best checkpoint | Step 3500 (by validation nDCG@10) |
| Hardware | NVIDIA RTX 3070 8GB, fp16 AMP |

---

## What is TalentCLEF Task B?

**TalentCLEF 2026 Task B** is a graded information-retrieval shared task:

- **Query**: a job title (e.g., *"Electrician"*)
- **Corpus**: 9,052 ESCO skills (e.g., *"install electric switches"*)
- **Relevance levels**: `2` = Core, `1` = Contextual, `0` = Non-relevant
- **Primary metric**: nDCG with graded relevance (core=2, contextual=1)

---

## Usage

### Installation

```bash

pip install sentence-transformers faiss-cpu

```

### Encode and Compare

```python

from sentence_transformers import SentenceTransformer



model = SentenceTransformer("talentguide/skillscout-large")



job    = "Data Scientist"

skills = ["data science", "machine learning", "install electric switches"]



embs   = model.encode([job] + skills, normalize_embeddings=True)

scores = embs[0] @ embs[1:].T



for skill, score in zip(skills, scores):

    print(f"{score:.3f}  {skill}")

# 0.872  data science

# 0.731  machine learning

# 0.112  install electric switches

```

### Full Retrieval with FAISS (Recommended)

```python

from sentence_transformers import SentenceTransformer

import faiss, numpy as np



model = SentenceTransformer("talentguide/skillscout-large")



# Build index once over your skill corpus

skill_texts = [...]  # list of skill names



embs = model.encode(skill_texts, batch_size=128,

                    normalize_embeddings=True,

                    show_progress_bar=True).astype(np.float32)



index = faiss.IndexFlatIP(embs.shape[1])  # inner product on L2-normed = cosine

index.add(embs)



job_title = "Software Engineer"

q = model.encode([job_title], normalize_embeddings=True).astype(np.float32)

scores, idxs = index.search(q, k=50)



for rank, (idx, score) in enumerate(zip(idxs[0], scores[0]), 1):

    print(f"{rank:3d}. [{score:.4f}]  {skill_texts[idx]}")

```

### Demo Output

```

Software Engineer

   1. [0.942]  define software architecture

   2. [0.938]  software frameworks

   3. [0.935]  create software design



Data Scientist

   1. [0.951]  data science

   2. [0.921]  establish data processes

   3. [0.919]  create data models



Electrician

   1. [0.944]  install electric switches

   2. [0.938]  install electricity sockets

   3. [0.930]  use electrical wire tools

```

---

## Two-Stage Pipeline Integration

```

Job title

   |

   v

[SkillScout Large]         <- this model

   |  top-200 candidates via FAISS ANN

   v

[Cross-encoder re-ranker]

   |  fine-grained re-scoring

   v

Final ranked list (graded: core > contextual > irrelevant)

```

Blend formula (alpha=0.7 gives best validation results):

```python

final_score = alpha * biencoder_score + (1 - alpha) * crossencoder_score

```

---

## Training Details

### Data

Source: [ESCO occupational ontology](https://esco.ec.europa.eu/), TalentCLEF 2026 training split.

| | Count |
|---|---|
| Job-skill pairs (essential) | ~57,500 |
| Job-skill pairs (optional) | ~57,200 |
| Total InputExamples | **93,720** |
| Validation queries | 304 |
| Validation corpus | 9,052 skills |
| Validation qrels | 56,417 |

Each ESCO job has 5-15 title aliases; skills have multiple phrasings.
Optional pairs are downsampled to 50% of essential count to maintain class balance.

### Hyperparameters

```

Loss           : MultipleNegativesRankingLoss (scale=20, cos_sim)

Batch size     : 64  (63 in-batch negatives per anchor)

Epochs         : 3

Warmup         : 10% of steps (~440 steps)

Optimizer      : AdamW fused

Learning rate  : 5e-5, linear decay

Precision      : fp16 AMP

Max seq len    : 64 tokens

Best model     : saved by cosine-nDCG@10 on validation

```

### Training Curve

| Epoch | Step | Train Loss | nDCG@10 val | MAP@100 val |
|---|---|---|---|---|
| 0.34 | 500  | 2.9232 | 0.3430 | - |
| 0.68 | 1000 | 2.1179 | 0.3424 | - |
| 1.00 | 1465 | -      | 0.3676 | 0.1758 |
| 1.37 | 2000 | 1.7070 | 0.3692 | - |
| 1.71 | 2500 | 1.6366 | 0.3744 | - |
| 2.00 | 2930 | -      | 0.3717 | 0.1780 |
| **2.39** | **3500** | **1.4540** | **0.3769** | **0.1808** |

### Validation Metrics (best checkpoint, step 3500)

| Metric | Value |
|---|---|
| **nDCG@10** | **0.4830** |
| nDCG@50 | 0.4240 |
| nDCG@100 | 0.3769 |
| **MAP@100** | **0.1825** |
| **MRR@10** | **0.6657** |
| Accuracy@1 | 0.5099 |
| Accuracy@3 | 0.7993 |
| Accuracy@5 | 0.8914 |
| Accuracy@10 | 0.9474 |

Evaluated with `InformationRetrievalEvaluator` (binary: any qrel > 0 = relevant).

### Pipeline Results (graded relevance, full 9052-skill ranking)

| Run | nDCG@10 graded | nDCG@10 binary | MAP |
|---|---|---|---|
| Zero-shot `jjzha/esco-xlm-roberta-large` | 0.2039 | 0.2853 | 0.2663 |
| **SkillScout Large (bi-encoder only)** | **0.3621** | **0.4830** | **0.4545** |
| SkillScout Large + cross-encoder (alpha=0.7) | **0.6896** | **0.7330** | 0.2481 |

---

## Competitive Context (TalentCLEF 2025 Task B)

| Team | MAP (test) | Approach |
|---|---|---|
| pjmathematician (winner 2025) | 0.36 | GTE 7B + contrastive + LLM-augmented data |
| NLPnorth (3rd of 14, 2025) | 0.29 | 3-class discriminative classification |
| **SkillScout Large (2026 val, Stage 1 only)** | **0.4545** | MNR fine-tuned bi-encoder |

---

## Limitations

- **English only** - trained on ESCO EN labels.
- **ESCO-domain optimised** - transfer to O*NET or custom taxonomies may require fine-tuning.

- **Max 64 tokens** - reduce long descriptions to a concise job title.

- **Graded distinction** - the bi-encoder alone does not reliably separate core vs contextual skills; a cross-encoder re-ranker is recommended for graded nDCG.



---



## Citation



```bibtex

@misc{talentguide-skillscout-2026,

  title  = {SkillScout Large: Dense Job-to-Skill Retrieval for TalentCLEF 2026},

  author = {TalentGuide},

  year   = {2026},

  url    = {https://huggingface.co/talentguide/skillscout-large}

}



@misc{talentclef2026taskb,

  title  = {TalentCLEF 2026 Task B: Job-Skill Matching},

  author = {TalentCLEF Organizers},

  year   = {2026},

  url    = {https://talentclef.github.io/}

}

```



---



## Framework Versions



- Python 3.12.10 | Sentence Transformers 5.3.0 | Transformers 5.5.0

- PyTorch 2.11.0+cu128 | Accelerate 1.13.0 | Tokenizers 0.22.2