Phase 2: reasoning SFT

updated 4 days ago

Reasoning SFT on 30K samples of OpenThoughts3 (OT30K) or another reasoning dataset, and the checkpoints each run starts from.

Upvote

ali-elganzory/1.7b-MixtureVitae-300BT-v1-decontaminated-16k

Text Generation • 2B • Updated Jul 27 • 17

Note MixtureVitae · starting point
ali-elganzory/1.7b-MixtureVitae-300BT-v1-decontaminated-16k-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated Jul 27 • 15

Note MixtureVitae · Tulu3 SFT, starting point
ali-elganzory/1.7b-MixtureVitae-300BT-v1-decontaminated-16k-DPO-Tulu3-decontaminated

Text Generation • 2B • Updated Jul 27 • 13

Note MixtureVitae · Tulu3 DPO, starting point
open-sci/sft__ot30k_1.7b-MixtureVitae-300BT-v1-decontaminated-16k

Feature Extraction • 2B • Updated 5 days ago • 14

Note MixtureVitae · OT30K
open-sci/sft__ot30k_1.7b-MixtureVitae-300BT-v1-decontaminated-16k-SFT-Tulu3-decontaminated

Feature Extraction • 2B • Updated May 3 • 9

Note MixtureVitae · Tulu3 SFT, then OT30K
open-sci/sft__ot30k_1.7b-MixtureVitae-300BT-v1-decontaminated-16k-DPO-Tulu3-decontaminated

Feature Extraction • 2B • Updated Apr 21 • 4

Note MixtureVitae · Tulu3 DPO, then OT30K
open-sci/1.7b-MixtureVitae-300BT-v1-decontaminated-16k-ar30k

Feature Extraction • 2B • Updated 5 days ago • 24

Note MixtureVitae · AceReason 30K
open-sci/1.7b-MixtureVitae-300BT-v1-decontaminated-16k-SFT-Tulu3-decontaminated-ar30k

Feature Extraction • 2B • Updated 5 days ago • 26

Note MixtureVitae · Tulu3 SFT, then AceReason 30K
open-sci/1.7b-MixtureVitae-300BT-v1-decontaminated-16k-DPO-Tulu3-decontaminated-ar30k

Feature Extraction • 2B • Updated 5 days ago • 25

Note MixtureVitae · Tulu3 DPO, then AceReason 30K
open-sci/1.7b-MixtureVitae-300BT-v1-decontaminated-16k-nc30k

Feature Extraction • 2B • Updated 5 days ago • 23

Note MixtureVitae · Nemotron-Cascade 30K
open-sci/1.7b-MixtureVitae-300BT-v1-decontaminated-16k-SFT-Tulu3-decontaminated-nc30k

Feature Extraction • 2B • Updated 5 days ago • 26

Note MixtureVitae · Tulu3 SFT, then Nemotron-Cascade 30K
open-sci/1.7b-MixtureVitae-300BT-v1-decontaminated-16k-DPO-Tulu3-decontaminated-nc30k

Feature Extraction • 2B • Updated 5 days ago • 23

Note MixtureVitae · Tulu3 DPO, then Nemotron-Cascade 30K
open-sci/1.7b-MixtureVitae-300BT-v1-decontaminated-16k-sr30k

Feature Extraction • 2B • Updated 5 days ago • 25

Note MixtureVitae · Superior-Reasoning 30K
open-sci/1.7b-MixtureVitae-300BT-v1-decontaminated-16k-SFT-Tulu3-decontaminated-sr30k

Feature Extraction • 2B • Updated 5 days ago • 26

Note MixtureVitae · Tulu3 SFT, then Superior-Reasoning 30K
open-sci/1.7b-MixtureVitae-300BT-v1-decontaminated-16k-DPO-Tulu3-decontaminated-sr30k

Feature Extraction • 2B • Updated 5 days ago • 23

Note MixtureVitae · Tulu3 DPO, then Superior-Reasoning 30K
ali-elganzory/1.7b-MixtureVitae-100BT-longsft_16k

Text Generation • 2B • Updated Jul 28 • 16 • 1

Note MixtureVitae, 100B tokens · starting point
ali-elganzory/1.7b-MixtureVitae-100BT-longsft_16k-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated Jul 28 • 24

Note MixtureVitae, 100B tokens · Tulu3 SFT, starting point
ali-elganzory/1.7b-MixtureVitae-100BT-longsft_16k-DPO-Tulu3-decontaminated

Text Generation • 2B • Updated Jul 29 • 22

Note MixtureVitae, 100B tokens · Tulu3 DPO, starting point
open-sci/1.7b-MixtureVitae-100BT-longsft_16k-ot30k

Feature Extraction • 2B • Updated 5 days ago • 29

Note MixtureVitae, 100B tokens · OT30K
open-sci/1.7b-MixtureVitae-100BT-longsft_16k-SFT-Tulu3-decontaminated-ot30k

Feature Extraction • 2B • Updated 5 days ago • 24

Note MixtureVitae, 100B tokens · Tulu3 SFT, then OT30K
open-sci/1.7b-MixtureVitae-100BT-longsft_16k-DPO-Tulu3-decontaminated-ot30k

Feature Extraction • 2B • Updated 5 days ago • 29

Note MixtureVitae, 100B tokens · Tulu3 DPO, then OT30K
laion/open-sci-ref-1.7b-mixturevitae-wo_ot3-100B-4096-longsft_16k

2B • Updated Jul 28 • 8

Note MixtureVitae without OT3, 100B tokens · starting point
ali-elganzory/open-sci-ref-1.7b-mixturevitae-wo_ot3-100B-4096-longsft_16k-SFT-Tulu3-decontaminated

Feature Extraction • 2B • Updated Jul 29 • 4

Note MixtureVitae without OT3, 100B tokens · Tulu3 SFT, starting point
open-sci/open-sci-ref-1.7b-mixturevitae-wo_ot3-100B-4096-longsft_16k-ot30k

Feature Extraction • 2B • Updated 5 days ago • 27

Note MixtureVitae without OT3, 100B tokens · OT30K
open-sci/open-sci-ref-1.7b-mixturevitae-wo_ot3-100B-4096-longsft_16k-SFT-Tulu3-decontaminated-ot30k

Feature Extraction • 2B • Updated 5 days ago • 24

Note MixtureVitae without OT3, 100B tokens · Tulu3 SFT, then OT30K
laion/open-sci-ref-1.7b-mixturevitae-wo_ot3-300B-4096-longsft_16k

2B • Updated Aug 1 • 8

Note MixtureVitae without OT3 · starting point
open-sci/open-sci-ref-1.7b-mixturevitae-wo_ot3-300B-4096-longsft_16k-ot30k

Feature Extraction • 2B • Updated 5 days ago • 26

Note MixtureVitae without OT3 · OT30K
ali-elganzory/open-sci-ref-v0.02-1.7b-nemotron-hq-300B-16k-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated Apr 8 • 17

Note Nemotron-CC-HQ, 16K pretraining run · Tulu3 SFT, starting point
open-sci/sft__ot30k_open-sci-ref-v0.02-1.7b-nemotron-hq-300B-16k-SFT-Tulu3-decontaminated

Feature Extraction • 2B • Updated 5 days ago • 21

Note Nemotron-CC-HQ, 16K pretraining run · Tulu3 SFT, then OT30K
open-sci/sft__ot30k_open-sci-ref-v0.02-1.7b-nemotron-hq-300B-16k-DPO-Tulu3-decontaminated

Feature Extraction • 2B • Updated Apr 21 • 6

Note Nemotron-CC-HQ, 16K pretraining run · Tulu3 DPO, then OT30K
open-sci/open-sci-ref-v0.02-1.7b-nemotron-hq-300B-16k-SFT-Tulu3-decontaminated-ar30k

Feature Extraction • 2B • Updated 5 days ago • 27

Note Nemotron-CC-HQ, 16K pretraining run · Tulu3 SFT, then AceReason 30K
open-sci/open-sci-ref-v0.02-1.7b-nemotron-hq-300B-16k-DPO-Tulu3-decontaminated-ar30k

Feature Extraction • 2B • Updated 5 days ago • 25

Note Nemotron-CC-HQ, 16K pretraining run · Tulu3 DPO, then AceReason 30K
open-sci/open-sci-ref-v0.02-1.7b-nemotron-hq-300B-16k-SFT-Tulu3-decontaminated-nc30k

Feature Extraction • 2B • Updated 5 days ago • 25

Note Nemotron-CC-HQ, 16K pretraining run · Tulu3 SFT, then Nemotron-Cascade 30K
open-sci/open-sci-ref-v0.02-1.7b-nemotron-hq-300B-16k-DPO-Tulu3-decontaminated-nc30k

Feature Extraction • 2B • Updated 5 days ago • 24

Note Nemotron-CC-HQ, 16K pretraining run · Tulu3 DPO, then Nemotron-Cascade 30K
open-sci/open-sci-ref-v0.02-1.7b-nemotron-hq-300B-16k-SFT-Tulu3-decontaminated-sr30k

Feature Extraction • 2B • Updated 5 days ago • 24

Note Nemotron-CC-HQ, 16K pretraining run · Tulu3 SFT, then Superior-Reasoning 30K
open-sci/open-sci-ref-v0.02-1.7b-nemotron-hq-300B-16k-DPO-Tulu3-decontaminated-sr30k

Feature Extraction • 2B • Updated 5 days ago • 25

Note Nemotron-CC-HQ, 16K pretraining run · Tulu3 DPO, then Superior-Reasoning 30K
ali-elganzory/open-sci-ref-1.7b-nemotron-hq-100B-4096-longsft_16k

2B • Updated Jul 28 • 8

Note Nemotron-CC-HQ, 100B tokens · starting point
ali-elganzory/open-sci-ref-v0.02-1.7b-nemotron-hq-100B-4096-longsft_16k-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated Jul 28 • 19

Note Nemotron-CC-HQ, 100B tokens · Tulu3 SFT, starting point
open-sci/open-sci-ref-1.7b-nemotron-hq-100B-4096-longsft_16k-ot30k

Feature Extraction • 2B • Updated 5 days ago • 23

Note Nemotron-CC-HQ, 100B tokens · OT30K
open-sci/open-sci-ref-v0.02-1.7b-nemotron-hq-100B-4096-longsft_16k-SFT-Tulu3-decontaminated-ot30k

Feature Extraction • 2B • Updated 5 days ago • 27

Note Nemotron-CC-HQ, 100B tokens · Tulu3 SFT, then OT30K
ali-elganzory/open-sci-ref-1.7b-nemotron-hq-mv_reasoning-100B-4096-longsft_16k

2B • Updated Jul 28 • 6

Note Nemotron-CC-HQ with MixtureVitae's reasoning subset, 100B tokens · starting point
ali-elganzory/open-sci-ref-v0.02-1.7b-nemotron-hq-mv_reasoning-100B-4096-longsft_16k-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated Jul 28 • 30

Note Nemotron-CC-HQ with MixtureVitae's reasoning subset, 100B tokens · Tulu3 SFT, starting point
open-sci/open-sci-ref-1.7b-nemotron-hq-mv_reasoning-100B-4096-longsft_16k-ot30k

Feature Extraction • 2B • Updated 5 days ago • 25

Note Nemotron-CC-HQ with MixtureVitae's reasoning subset, 100B tokens · OT30K
ali-elganzory/1.7b-Comma0.1-300BT-longsft_16k-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated Apr 13 • 29

Note Comma0.1 · Tulu3 SFT, starting point
open-sci/sft__ot30k_1.7b-Comma0.1-300BT-longsft_16k-SFT-Tulu3-decontaminated

Feature Extraction • 2B • Updated Apr 21 • 3

Note Comma0.1 · Tulu3 SFT, then OT30K
open-sci/sft__ot30k_1.7b-Comma0.1-300BT-longsft_16k-DPO-Tulu3-decontaminate

Feature Extraction • 2B • Updated 5 days ago • 18

Note Comma0.1 · Tulu3 DPO, then OT30K
open-sci/1.7b-Comma0.1-300BT-longsft_16k-SFT-Tulu3-decontaminated-ar30k

Feature Extraction • 2B • Updated 5 days ago • 22

Note Comma0.1 · Tulu3 SFT, then AceReason 30K
open-sci/1.7b-Comma0.1-300BT-longsft_16k-DPO-Tulu3-decontaminated-ar30k

Feature Extraction • 2B • Updated 5 days ago • 22

Note Comma0.1 · Tulu3 DPO, then AceReason 30K
open-sci/1.7b-Comma0.1-300BT-longsft_16k-SFT-Tulu3-decontaminated-nc30k

Feature Extraction • 2B • Updated 5 days ago • 23

Note Comma0.1 · Tulu3 SFT, then Nemotron-Cascade 30K
open-sci/1.7b-Comma0.1-300BT-longsft_16k-DPO-Tulu3-decontaminated-nc30k

Feature Extraction • 2B • Updated 5 days ago • 22

Note Comma0.1 · Tulu3 DPO, then Nemotron-Cascade 30K
open-sci/1.7b-Comma0.1-300BT-longsft_16k-SFT-Tulu3-decontaminated-sr30k

Feature Extraction • 2B • Updated 5 days ago • 22

Note Comma0.1 · Tulu3 SFT, then Superior-Reasoning 30K
open-sci/1.7b-Comma0.1-300BT-longsft_16k-DPO-Tulu3-decontaminated-sr30k

Feature Extraction • 2B • Updated 5 days ago • 19

Note Comma0.1 · Tulu3 DPO, then Superior-Reasoning 30K
ali-elganzory/open-sci-ref-v0.02-1.7b-fineweb-edu-1.4t-300B-4096-longsft_16k-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated Jul 28 • 32

Note FineWeb-Edu · Tulu3 SFT, starting point
open-sci/sft__ot30k_open-sci-ref-v0.02-1.7b-fineweb-edu-1.4t-300B-4096-longsft_16k-SFT-Tulu3

Feature Extraction • 2B • Updated Apr 21 • 5

Note FineWeb-Edu · Tulu3 SFT, then OT30K
open-sci/open-sci-ref-v0.02-1.7b-fineweb-edu-1.4t-300B-4096-longsft_16k-SFT-Tulu3-decontaminated-ar30k

Feature Extraction • 2B • Updated 5 days ago • 25

Note FineWeb-Edu · Tulu3 SFT, then AceReason 30K
open-sci/open-sci-ref-v0.02-1.7b-fineweb-edu-1.4t-300B-4096-longsft_16k-DPO-Tulu3-decontaminated-ar30k

Feature Extraction • 2B • Updated 5 days ago • 23

Note FineWeb-Edu · Tulu3 DPO, then AceReason 30K
open-sci/open-sci-ref-v0.02-1.7b-fineweb-edu-1.4t-300B-4096-longsft_16k-SFT-Tulu3-decontaminated-nc30k

Feature Extraction • 2B • Updated 5 days ago • 24

Note FineWeb-Edu · Tulu3 SFT, then Nemotron-Cascade 30K
open-sci/open-sci-ref-v0.02-1.7b-fineweb-edu-1.4t-300B-4096-longsft_16k-DPO-Tulu3-decontaminated-nc30k

Feature Extraction • 2B • Updated 5 days ago • 25

Note FineWeb-Edu · Tulu3 DPO, then Nemotron-Cascade 30K
open-sci/open-sci-ref-v0.02-1.7b-fineweb-edu-1.4t-300B-4096-longsft_16k-SFT-Tulu3-decontaminated-sr30k

Feature Extraction • 2B • Updated 5 days ago • 23

Note FineWeb-Edu · Tulu3 SFT, then Superior-Reasoning 30K
open-sci/open-sci-ref-v0.02-1.7b-fineweb-edu-1.4t-300B-4096-longsft_16k-DPO-Tulu3-decontaminated-sr30k

Feature Extraction • 2B • Updated 5 days ago • 24

Note FineWeb-Edu · Tulu3 DPO, then Superior-Reasoning 30K
ali-elganzory/open-sci-ref-v0.02-1.7b-dclm-300B-4096-longsft_16k-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated 8 days ago • 378

Note DCLM · Tulu3 SFT, starting point
open-sci/open-sci-ref-v0.02-1.7b-dclm-300B-4096-longsft_16k-SFT-Tulu3-decontaminate-ot30k

Feature Extraction • 2B • Updated 5 days ago • 26

Note DCLM · Tulu3 SFT, then OT30K
open-sci/open-sci-ref-v0.02-1.7b-dclm-300B-4096-longsft_16k-SFT-Tulu3-decontaminate-ot30k-rope8-32k

Feature Extraction • 2B • Updated 5 days ago • 27

Note DCLM · Tulu3 SFT, then OT30K
ali-elganzory/SmolLM2-1.7B-16k-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated Apr 13 • 37

Note SmolLM2-1.7B · Tulu3 SFT, starting point
open-sci/sft__ot30k_SmolLM2-1.7B-16k-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated 5 days ago • 205

Note SmolLM2-1.7B · Tulu3 SFT, then OT30K
open-sci/SmolLM2-1.7B-16k-SFT-Tulu3-decontaminated-ar30k

Text Generation • 2B • Updated 5 days ago • 212

Note SmolLM2-1.7B · Tulu3 SFT, then AceReason 30K
open-sci/SmolLM2-1.7B-16k-DPO-Tulu3-decontaminated-ar30k

Text Generation • 2B • Updated 5 days ago • 191

Note SmolLM2-1.7B · Tulu3 DPO, then AceReason 30K
open-sci/SmolLM2-1.7B-16k-SFT-Tulu3-decontaminated-nc30k

Text Generation • 2B • Updated 5 days ago • 211

Note SmolLM2-1.7B · Tulu3 SFT, then Nemotron-Cascade 30K
open-sci/SmolLM2-1.7B-16k-DPO-Tulu3-decontaminated-nc30k

Text Generation • 2B • Updated 5 days ago • 203

Note SmolLM2-1.7B · Tulu3 DPO, then Nemotron-Cascade 30K
open-sci/SmolLM2-1.7B-16k-SFT-Tulu3-decontaminated-sr30k

Text Generation • 2B • Updated 5 days ago • 209

Note SmolLM2-1.7B · Tulu3 SFT, then Superior-Reasoning 30K
open-sci/SmolLM2-1.7B-16k-DPO-Tulu3-decontaminated-sr30k

Text Generation • 2B • Updated 5 days ago • 204

Note SmolLM2-1.7B · Tulu3 DPO, then Superior-Reasoning 30K
HuggingFaceTB/SmolLM2-1.7B-Instruct-16k

Text Generation • 2B • Updated Feb 21, 2025 • 1.25k • 10

Note SmolLM2-1.7B-Instruct · starting point
open-sci/sft__ot30k_SmolLM2-1.7B-Instruct-16k

Text Generation • 2B • Updated 5 days ago • 193

Note SmolLM2-1.7B-Instruct · OT30K
ali-elganzory/Qwen2.5-1.5B-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated Jan 18 • 79

Note Qwen2.5-1.5B · Tulu3 SFT, starting point
open-sci/sft__ot30k_Qwen2.5-1.5B-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated Apr 21 • 33

Note Qwen2.5-1.5B · Tulu3 SFT, then OT30K
open-sci/sft__ot30k_Qwen2.5-1.5B-DPO-Tulu3-decontaminated

Text Generation • 2B • Updated Apr 21 • 32

Note Qwen2.5-1.5B · Tulu3 DPO, then OT30K
open-sci/Qwen2.5-1.5B-SFT-Tulu3-decontaminated-ar30k

Text Generation • 2B • Updated 5 days ago • 173

Note Qwen2.5-1.5B · Tulu3 SFT, then AceReason 30K
open-sci/Qwen2.5-1.5B-DPO-Tulu3-decontaminated-ar30k

Text Generation • 2B • Updated 5 days ago • 172

Note Qwen2.5-1.5B · Tulu3 DPO, then AceReason 30K
open-sci/Qwen2.5-1.5B-SFT-Tulu3-decontaminated-nc30k

Text Generation • 2B • Updated 5 days ago • 174

Note Qwen2.5-1.5B · Tulu3 SFT, then Nemotron-Cascade 30K
open-sci/Qwen2.5-1.5B-DPO-Tulu3-decontaminated-nc30k

Text Generation • 2B • Updated 5 days ago • 174

Note Qwen2.5-1.5B · Tulu3 DPO, then Nemotron-Cascade 30K
open-sci/Qwen2.5-1.5B-SFT-Tulu3-decontaminated-sr30k

Text Generation • 2B • Updated 5 days ago • 175

Note Qwen2.5-1.5B · Tulu3 SFT, then Superior-Reasoning 30K
open-sci/Qwen2.5-1.5B-DPO-Tulu3-decontaminated-sr30k

Text Generation • 2B • Updated 5 days ago • 170

Note Qwen2.5-1.5B · Tulu3 DPO, then Superior-Reasoning 30K
Qwen/Qwen2.5-1.5B-Instruct

Text Generation • 2B • Updated Sep 25, 2024 • 4.68M • • 929

Note Qwen2.5-1.5B-Instruct · starting point
open-sci/sft__ot30k_Qwen2.5-1.5B

Text Generation • 2B • Updated 5 days ago • 191

Note Qwen2.5-1.5B-Instruct · OT30K
ali-elganzory/Qwen3-1.7B-Base-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated Jan 18 • 49

Note Qwen3-1.7B-Base · Tulu3 SFT, starting point
open-sci/sft__ot30k_Qwen3-1.7B-Base-SFT-Tulu3-decontaminated

Text Generation • 2B • Updated Apr 21 • 33

Note Qwen3-1.7B-Base · Tulu3 SFT, then OT30K
open-sci/sft__ot30k_Qwen3-1.7B-Base-DPO-Tulu3-decontaminated

Text Generation • 2B • Updated Apr 21 • 18

Note Qwen3-1.7B-Base · Tulu3 DPO, then OT30K
open-sci/Qwen3-1.7B-Base-SFT-Tulu3-decontaminated-ar30k

Text Generation • 2B • Updated 5 days ago • 174

Note Qwen3-1.7B-Base · Tulu3 SFT, then AceReason 30K
open-sci/Qwen3-1.7B-Base-DPO-Tulu3-decontaminated-ar30k

Text Generation • 2B • Updated 5 days ago • 168

Note Qwen3-1.7B-Base · Tulu3 DPO, then AceReason 30K
open-sci/Qwen3-1.7B-Base-SFT-Tulu3-decontaminated-nc30k

Text Generation • 2B • Updated 5 days ago • 180

Note Qwen3-1.7B-Base · Tulu3 SFT, then Nemotron-Cascade 30K
open-sci/Qwen3-1.7B-Base-DPO-Tulu3-decontaminated-nc30k

Text Generation • 2B • Updated 5 days ago • 171

Note Qwen3-1.7B-Base · Tulu3 DPO, then Nemotron-Cascade 30K
open-sci/Qwen3-1.7B-Base-SFT-Tulu3-decontaminated-sr30k

Text Generation • 2B • Updated 5 days ago • 191

Note Qwen3-1.7B-Base · Tulu3 SFT, then Superior-Reasoning 30K
open-sci/Qwen3-1.7B-Base-DPO-Tulu3-decontaminated-sr30k

Text Generation • 2B • Updated 5 days ago • 169

Note Qwen3-1.7B-Base · Tulu3 DPO, then Superior-Reasoning 30K
Qwen/Qwen3-1.7B

Text Generation • 2B • Updated Jul 26, 2025 • 4.31M • • 613

Note Qwen3-1.7B · starting point
open-sci/sft__ot30k_Qwen3-1.7B-Base

Text Generation • 2B • Updated 5 days ago • 194

Note Qwen3-1.7B · OT30K
permissive-post-train/evaluation-records

Updated 3 days ago • 216

Note The evaluation records of every checkpoint here

Upvote