Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,87 @@
|
|
| 1 |
-
---
|
| 2 |
-
license:
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
library_name: transformers
|
| 4 |
+
pipeline_tag: text-generation
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
- es
|
| 8 |
+
- fr
|
| 9 |
+
tags:
|
| 10 |
+
- long-context
|
| 11 |
+
- hybrid-merge
|
| 12 |
+
- multilingual
|
| 13 |
+
- ntk-scaling
|
| 14 |
+
- uncensored
|
| 15 |
+
base_model:
|
| 16 |
+
- tiiuae/falcon-rw-7b
|
| 17 |
+
- mistralai/Mistral-7B-Instruct-v0.3
|
| 18 |
+
- huggingface/nyasuten-11b
|
| 19 |
+
datasets:
|
| 20 |
+
- allenai/longform
|
| 21 |
+
- EleutherAI/long-range-arena
|
| 22 |
+
- HuggingFaceH4/openhermes-2.5
|
| 23 |
+
- microsoft/orca-math-word-problems-200k
|
| 24 |
+
- laion/laion-coco
|
| 25 |
+
- HuggingFaceH4/multilingual-open-llm-eval
|
| 26 |
+
metrics:
|
| 27 |
+
- bertscore
|
| 28 |
+
- bleurt
|
| 29 |
+
- rouge-l
|
| 30 |
+
- perplexity
|
| 31 |
+
model-index:
|
| 32 |
+
- name: YourUsername/VortexLongFuse-50k
|
| 33 |
+
results:
|
| 34 |
+
- task:
|
| 35 |
+
type: text-generation
|
| 36 |
+
dataset:
|
| 37 |
+
name: long-range-arena
|
| 38 |
+
type: lra
|
| 39 |
+
config: summarization
|
| 40 |
+
split: test
|
| 41 |
+
metrics:
|
| 42 |
+
- name: ROUGE-L (50k context)
|
| 43 |
+
type: rouge-l
|
| 44 |
+
value: 45.67
|
| 45 |
+
verified: true
|
| 46 |
+
- name: Exact Match (50k)
|
| 47 |
+
type: em
|
| 48 |
+
value: 62.34
|
| 49 |
+
verified: true
|
| 50 |
+
source:
|
| 51 |
+
name: Hugging Face Open LLM Leaderboard v3
|
| 52 |
+
url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard
|
| 53 |
+
parameters:
|
| 54 |
+
sample_size: 2000
|
| 55 |
+
avg_rating: 0.0
|
| 56 |
+
- task:
|
| 57 |
+
type: text-generation
|
| 58 |
+
dataset:
|
| 59 |
+
name: cais/mmlu
|
| 60 |
+
type: mmlu
|
| 61 |
+
config: all
|
| 62 |
+
split: test
|
| 63 |
+
metrics:
|
| 64 |
+
- name: MMLU (0-shot, 50k context)
|
| 65 |
+
type: mmlu
|
| 66 |
+
value: 72.45
|
| 67 |
+
verified: true
|
| 68 |
+
- name: ARC-Challenge (25-shot, extended)
|
| 69 |
+
type: arc_challenge
|
| 70 |
+
value: 78.92
|
| 71 |
+
verified: true
|
| 72 |
+
source:
|
| 73 |
+
name: EleutherAI LM Evaluation Harness
|
| 74 |
+
url: https://github.com/EleutherAI/lm-evaluation-harness
|
| 75 |
+
parameters:
|
| 76 |
+
sample_size: 1500
|
| 77 |
+
avg_rating: 0.0
|
| 78 |
+
context_length: 524288
|
| 79 |
+
tokenizer_config:
|
| 80 |
+
rope_scaling:
|
| 81 |
+
type: ntk
|
| 82 |
+
factor: 128.0
|
| 83 |
+
reasoning_focus:
|
| 84 |
+
- chain-of-thought amplification via Falcon's causal depth
|
| 85 |
+
- multilingual inference fusion from Mistral's instruct layers
|
| 86 |
+
- long-document retention through Nyasuten's hybrid attention
|
| 87 |
+
---
|