Downloads · 30 days
15
4% of all-time downloads
HyperCLOVA-X-SEED-Think-14B-correct-only-r1-ties
youngseok12/HyperCLOVA-X-SEED-Think-14B-correct-only-r1-ties
HyperCLOVA-X-SEED-Think-14B-correct-only-r1-ties is a text generation model from youngseok12. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as other.
naver-hyperclovax/HyperCLOVAX-SEED-Think-14B pristine Base에서 세 AI Hub source를 각각 독립적으로 LoRA SFT한 뒤 TIES로 병합한 standalone BF16 모델입니다. 각 source의 1,000개 Base probe 중 Base가 정답을 맞힌 sample만 사용하여, source별 correct-only 학습 효과와…
All-time downloads
380
Public
Parameters
14.7B
29.5 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors29.5 GB · 100%
At a glance
- Task
- Text Generation
- Library
- transformers
- License
- other
- Model type
- hyperclovax
- Access
- Public
- Created
- Sep 6, 2026
- Updated
- Sep 6, 2026
- SHA
- 08487dbe
Base models
Model card
From the Hugging Face model README
HyperCLOVA X SEED Think-14B Correct-Only TIES
naver-hyperclovax/HyperCLOVAX-SEED-Think-14B pristine Base에서 세 AI Hub source를
각각 독립적으로 LoRA SFT한 뒤 TIES로 병합한 standalone BF16 모델입니다. 각 source의
1,000개 Base probe 중 Base가 정답을 맞힌 sample만 사용하여, source별 correct-only
학습 효과와 세 specialist의 병합 효과를 확인하기 위한 연구·평가용 모델입니다.
Model information
- Base model:
naver-hyperclovax/HyperCLOVAX-SEED-Think-14B - Base revision:
9b74e35d4c7e4ffec489f4171273caca8948a2b9 - Format: standalone BF16
safetensors; 별도 adapter가 필요하지 않습니다. - Merge: TIES, weights
[1.0, 1.0, 1.0], density0.5 - Chat template: Base 원본 그대로 사용했으며 수정하지 않았습니다.
Training data
세 adapter는 모두 pristine Base에서 독립적으로 학습했으며, 서로의 학습 데이터나 기존 mixture를 섞지 않았습니다. 아래 수는 각 1,000개 probe에서 correct로 판정된 실제 학습 sample 수입니다.
| Specialist | AI Hub dataset | Correct-only samples | Dataset URL |
|---|---|---|---|
| K | 71875 필수의료 의학지식 QA | 557 | AI Hub 71875 |
| R | 71568 경제·스포츠 숫자 연산/수치 MRC | 131 | AI Hub 71568 |
| C | 71949 인과관계 기반 추론 | 535 | AI Hub 71949 |
총 1,223개 sample에 해당하는 세 correct-only adapter를 병합했습니다. 학습에는 public benchmark test 데이터, v0.21 mixture, 다른 fine-tuned checkpoint를 사용하지 않았습니다. 각 target은 기존 answer-first 형식을 유지했습니다.
Training configuration
| Method | independent LoRA SFT, then TIES merge |
| Rank / alpha / dropout | 16 / 32 / 0.05 |
| Target modules | q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj |
| Learning rate | 5e-5, cosine scheduler, warmup ratio 0.03 |
| Epochs | 1 |
| Batch | per-device 4, gradient accumulation 4, effective 16 |
| Sequence length | 4096, packing disabled |
| Precision | BF16 |
| Objective | assistant-token-only causal LM cross entropy |
| Seeds | training/data 42/42 |
| Weight decay | 0 |
Validation
The merged weights passed a full NaN/Inf sweep over 421 parameters. The model was loaded and generated responses for 24 canonical smoke samples with zero empty outputs. This is a load/generation smoke check; no benchmark score is reported in this repository.
License and use
This derivative model is distributed under the Base model's HyperCLOVA X SEED 14B Think
Model License Agreement included in LICENSE. Please review that agreement and the
applicable AI Hub dataset terms before use or redistribution. The model is for research
- Task
- Text Generation
- Library
- transformers
- Type
- hyperclovax
- License
- other
- Languages
- ko
- Created
- Sep 6, 2026
- Updated
- Sep 6, 2026