File size: 11,047 Bytes
84a54bc |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 |
---
language:
- en
license: apache-2.0
library_name: llm2ner
base_model: meta-llama/Llama-3.2-1B
tags:
- ner
- span-detection
- llm
- pytorch
pipeline_tag: token-classification
model_name: ToMMeR-Llama-3.2-1B_L2_R64
source: https://github.com/VictorMorand/llm2ner
paper: https://arxiv.org/abs/2510.19410
---
# ToMMeR-Llama-3.2-1B_L2_R64
ToMMeR is a lightweight probing model extracting emergent mention detection capabilities from early layers representations of any LLM backbone, achieving high Zero Shot recall across a wide set of 13 NER benchmarks.
## Checkpoint Details
| Property | Value |
|-----------|-------|
| Base LLM | `meta-llama/Llama-3.2-1B` |
| Layer | 2|
| #Params | 264.2K |
# Usage
## Installation
Our code can be installed with pip+git, Please visit the [repository](https://github.com/VictorMorand/llm2ner) for more details.
```bash
pip install git+https://github.com/VictorMorand/llm2ner.git
```
## Fancy Outputs
```python
import llm2ner
from llm2ner import ToMMeR
tommer = ToMMeR.from_pretrained("llm2ner/ToMMeR-Llama-3.2-1B_L2_R64")
# load Backbone llm, optionnally cut the unused layer to save GPU space.
llm = llm2ner.utils.load_llm( tommer.llm_name, cut_to_layer=tommer.layer,)
tommer.to(llm.device)
text = "Large language models are awesome. While trained on language modeling, they exhibit emergent Zero Shot abilities that make them suitable for a wide range of tasks, including Named Entity Recognition (NER). "
#fancy interactive output
outputs = llm2ner.plotting.demo_inference( text, tommer, llm,
decoding_strategy="threshold", # or "greedy" for flat segmentation
threshold=0.5, # default 50%
show_attn=True,
)
```
<div>
<span class="tex2jax_ignore"><div class="spans" style="line-height: 2.5; direction: ltr">
<span style="font-weight: bold; display: inline-block; position: relative; height: 60px;">
Large
<span style="background: lightblue; top: 40px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
<span style="background: lightblue; top: 40px; height: 4px; border-top-left-radius: 3px; border-bottom-left-radius: 3px; left: -1px; width: calc(100% + 2px); position: absolute;">
<span style="background: lightblue; z-index: 10; color: #000; top: -0.5em; padding: 2px 3px; position: absolute; font-size: 0.6em; font-weight: bold; line-height: 1; border-radius: 3px">
PRED
</span>
</span>
</span>
<span style="font-weight: bold; display: inline-block; position: relative; height: 77px;">
language
<span style="background: lightblue; top: 40px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
<span style="background: lightblue; top: 57px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
<span style="background: lightblue; top: 57px; height: 4px; border-top-left-radius: 3px; border-bottom-left-radius: 3px; left: -1px; width: calc(100% + 2px); position: absolute;">
<span style="background: lightblue; z-index: 10; color: #000; top: -0.5em; padding: 2px 3px; position: absolute; font-size: 0.6em; font-weight: bold; line-height: 1; border-radius: 3px">
PRED
</span>
</span>
</span>
<span style="font-weight: bold; display: inline-block; position: relative; height: 77px;">
models
<span style="background: lightblue; top: 40px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
<span style="background: lightblue; top: 57px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
</span>
are awesome . While trained on
<span style="font-weight: bold; display: inline-block; position: relative; height: 60px;">
language
<span style="background: lightblue; top: 40px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
<span style="background: lightblue; top: 40px; height: 4px; border-top-left-radius: 3px; border-bottom-left-radius: 3px; left: -1px; width: calc(100% + 2px); position: absolute;">
<span style="background: lightblue; z-index: 10; color: #000; top: -0.5em; padding: 2px 3px; position: absolute; font-size: 0.6em; font-weight: bold; line-height: 1; border-radius: 3px">
PRED
</span>
</span>
</span>
<span style="font-weight: bold; display: inline-block; position: relative; height: 60px;">
modeling
<span style="background: lightblue; top: 40px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
</span>
, they exhibit
<span style="font-weight: bold; display: inline-block; position: relative; height: 60px;">
emergent
<span style="background: lightblue; top: 40px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
<span style="background: lightblue; top: 40px; height: 4px; border-top-left-radius: 3px; border-bottom-left-radius: 3px; left: -1px; width: calc(100% + 2px); position: absolute;">
<span style="background: lightblue; z-index: 10; color: #000; top: -0.5em; padding: 2px 3px; position: absolute; font-size: 0.6em; font-weight: bold; line-height: 1; border-radius: 3px">
PRED
</span>
</span>
</span>
<span style="font-weight: bold; display: inline-block; position: relative; height: 60px;">
abilities
<span style="background: lightblue; top: 40px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
</span>
that make them suitable for a wide range of
<span style="font-weight: bold; display: inline-block; position: relative; height: 60px;">
tasks
<span style="background: lightblue; top: 40px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
<span style="background: lightblue; top: 40px; height: 4px; border-top-left-radius: 3px; border-bottom-left-radius: 3px; left: -1px; width: calc(100% + 2px); position: absolute;">
<span style="background: lightblue; z-index: 10; color: #000; top: -0.5em; padding: 2px 3px; position: absolute; font-size: 0.6em; font-weight: bold; line-height: 1; border-radius: 3px">
PRED
</span>
</span>
</span>
, including
<span style="font-weight: bold; display: inline-block; position: relative; height: 60px;">
Named
<span style="background: lightblue; top: 40px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
<span style="background: lightblue; top: 40px; height: 4px; border-top-left-radius: 3px; border-bottom-left-radius: 3px; left: -1px; width: calc(100% + 2px); position: absolute;">
<span style="background: lightblue; z-index: 10; color: #000; top: -0.5em; padding: 2px 3px; position: absolute; font-size: 0.6em; font-weight: bold; line-height: 1; border-radius: 3px">
PRED
</span>
</span>
</span>
<span style="font-weight: bold; display: inline-block; position: relative; height: 60px;">
Entity
<span style="background: lightblue; top: 40px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
</span>
<span style="font-weight: bold; display: inline-block; position: relative; height: 60px;">
Recognition
<span style="background: lightblue; top: 40px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
</span>
(
<span style="font-weight: bold; display: inline-block; position: relative; height: 60px;">
NER
<span style="background: lightblue; top: 40px; height: 4px; left: -1px; width: calc(100% + 2px); position: absolute;">
</span>
<span style="background: lightblue; top: 40px; height: 4px; border-top-left-radius: 3px; border-bottom-left-radius: 3px; left: -1px; width: calc(100% + 2px); position: absolute;">
<span style="background: lightblue; z-index: 10; color: #000; top: -0.5em; padding: 2px 3px; position: absolute; font-size: 0.6em; font-weight: bold; line-height: 1; border-radius: 3px">
PRED
</span>
</span>
</span>
) . </div></span>
</div>
## Raw inference
By default, ToMMeR outputs span probabilities, but we also propose built-in options for decoding entities.
- Inputs:
- tokens (batch, seq): tokens to process,
- model: LLM to extract representation from.
- Outputs: (batch, seq, seq) matrix (masked outside valid spans)
```python
tommer = ToMMeR.from_pretrained("llm2ner/ToMMeR-Llama-3.2-1B_L2_R64")
# load Backbone llm, optionnally cut the unused layer to save GPU space.
llm = llm2ner.utils.load_llm( tommer.llm_name, cut_to_layer=tommer.layer,)
tommer.to(llm.device)
#### Raw Inference
text = ["Large language models are awesome"]
print(f"Input text: {text[0]}")
#tokenize in shape (1, seq_len)
tokens = model.tokenizer(text, return_tensors="pt")["input_ids"].to(device)
# Output raw scores
output = tommer.forward(tokens, model) # (batch_size, seq_len, seq_len)
print(f"Raw Output shape: {output.shape}")
#use given decoding strategy to infer entities
entities = tommer.infer_entities(tokens=tokens, model=model, threshold=0.5, decoding_strategy="greedy")
str_entities = [ model.tokenizer.decode(tokens[0,b:e+1]) for b, e in entities[0]]
print(f"Predicted entities: {str_entities}")
>>> Input text: Large language models are awesome
>>> Raw Output shape: torch.Size([1, 6, 6])
>>> Predicted entities: ['Large language models']
```
Please visit the [repository](https://github.com/VictorMorand/llm2ner) for more details and a demo notebook.
## Evaluation Results
| dataset | precision | recall | f1 | n_samples |
|---------------------|-------------|----------|--------|-------------|
| MultiNERD | 0.1668 | 0.992 | 0.2855 | 154144 |
| CoNLL 2003 | 0.2409 | 0.9643 | 0.3856 | 16493 |
| CrossNER_politics | 0.2464 | 0.9762 | 0.3935 | 1389 |
| CrossNER_AI | 0.2506 | 0.9749 | 0.3988 | 879 |
| CrossNER_literature | 0.2688 | 0.9592 | 0.4199 | 916 |
| CrossNER_science | 0.2774 | 0.9727 | 0.4317 | 1193 |
| CrossNER_music | 0.3037 | 0.9686 | 0.4625 | 945 |
| ncbi | 0.0964 | 0.9407 | 0.1748 | 3952 |
| FabNER | 0.2626 | 0.8111 | 0.3967 | 13681 |
| WikiNeural | 0.1609 | 0.9911 | 0.2769 | 92672 |
| GENIA_NER | 0.1886 | 0.9696 | 0.3157 | 16563 |
| ACE 2005 | 0.2514 | 0.4976 | 0.334 | 8230 |
| Ontonotes | 0.2015 | 0.7659 | 0.319 | 42193 |
| Aggregated | 0.1803 | 0.943 | 0.3027 | 353250 |
| Mean | 0.2243 | 0.9065 | 0.3534 | 353250 |
## Citation
If using this model or the approach, please cite the associated paper:
```
@misc{morand2025tommerefficiententity,
title={ToMMeR -- Efficient Entity Mention Detection from Large Language Models},
author={Victor Morand and Nadi Tomeh and Josiane Mothe and Benjamin Piwowarski},
year={2025},
eprint={2510.19410},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2510.19410},
}
```
## License
Apache-2.0 (see repository for full text). |