TimeCrystal-l2-13B / README.md
leaderboard-pr-bot's picture
Adding Evaluation Results
25bd7ee verified
|
raw
history blame
4.15 kB
---
license: apache-2.0
tags:
- llama-2
- roleplaying
model-index:
- name: TimeCrystal-l2-13B
results:
- task:
type: text-generation
name: Text Generation
dataset:
name: AI2 Reasoning Challenge (25-Shot)
type: ai2_arc
config: ARC-Challenge
split: test
args:
num_few_shot: 25
metrics:
- type: acc_norm
value: 61.18
name: normalized accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=BlueNipples/TimeCrystal-l2-13B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: HellaSwag (10-Shot)
type: hellaswag
split: validation
args:
num_few_shot: 10
metrics:
- type: acc_norm
value: 83.71
name: normalized accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=BlueNipples/TimeCrystal-l2-13B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MMLU (5-Shot)
type: cais/mmlu
config: all
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 56.46
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=BlueNipples/TimeCrystal-l2-13B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: TruthfulQA (0-shot)
type: truthful_qa
config: multiple_choice
split: validation
args:
num_few_shot: 0
metrics:
- type: mc2
value: 51.3
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=BlueNipples/TimeCrystal-l2-13B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: Winogrande (5-shot)
type: winogrande
config: winogrande_xl
split: validation
args:
num_few_shot: 5
metrics:
- type: acc
value: 75.37
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=BlueNipples/TimeCrystal-l2-13B
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: GSM8k (5-shot)
type: gsm8k
config: main
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 27.52
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=BlueNipples/TimeCrystal-l2-13B
name: Open LLM Leaderboard
---
This 13B model, TimeCrystal-l2-13B is built to maximize logic and instruct following, whilst also increasing the vividness of prose found in Chronos based models like Mythomax, over the more romantic prose, hopefully without losing the elegent narrative structure touch of newer models like synthia and xwin. TLDR: Attempt at more clever, better prose.
Tentative test results: I'm not certain if logic/instruct was improved or not (haven't tested much), but the prose infusion seems to have worked really well.
It is built so:
SLERPS:
Amethyst + Openchat Super = OpenStone
MythoMax + Chronos = ChronoMax
ChronoMax + Amethyst = TimeStone
Gradient Merge:
TimeStone + OpenStone (0.9,0,0) = TimeCrystal
Props to all the mergers, fine tuners!
All models in Merge: Many, lol.
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_BlueNipples__TimeCrystal-l2-13B)
| Metric |Value|
|---------------------------------|----:|
|Avg. |59.26|
|AI2 Reasoning Challenge (25-Shot)|61.18|
|HellaSwag (10-Shot) |83.71|
|MMLU (5-Shot) |56.46|
|TruthfulQA (0-shot) |51.30|
|Winogrande (5-shot) |75.37|
|GSM8k (5-shot) |27.52|