paper-with-me

Language Modelling 벤치마크

Language Modelling on SALMon

10개 결과 · ⬇ CSV · JSON

Sentiment Consistency 낮을수록 좋음

40.5 48.75 57 65.25 73.5 2021-09 2026-09 pGSLM — 40.5 (2021-09-07) TWIST 7B — 61.5 (2023-05-22) TWIST 1.3B — 61.5 (2023-05-22) TWIST 350M — 59.0 (2023-05-22) Spirit-LM (Expr.) — 73.5 (2024-02-08) Spirit-LM (base) — 54.5 (2024-02-08) LAST 1.3B — 65.0 (2024-09-05) LAST 350M — 64.0 (2024-09-05) pGSLM — 40.5 (2021-09-07)
RankModel Sentiment ConsistencySpeaker ConsistencyGender ConsistencyBackground (Domain) ConsistencyBackground (Random) ConsistencyRoom ConsistencySentiment AlignmentBackground Alignment Extra Training Data PaperCodeYear
1 Spirit-LM (Expr.) 73.581.085.055.064.054.552.059.5 Spirit LM: Interleaved Spoken and Written Language Model facebookresearch/spiritlm 2024
2 LAST 1.3B 65.064.568.556.061.062.553.553.0 LAST: Language Model Aware Speech Tokenization 2024
3 LAST 350M 64.063.070.555.560.561.051.554.5 LAST: Language Model Aware Speech Tokenization 2024
4 TWIST 7B 61.571.070.055.060.562.051.554.5 Textually Pretrained Speech Language Models slp-rl/spokenstorycloze 2023
4 TWIST 1.3B 61.569.069.555.560.559.053.056.5 Textually Pretrained Speech Language Models slp-rl/spokenstorycloze 2023
6 TWIST 350M 59.069.568.054.061.559.051.556.5 Textually Pretrained Speech Language Models slp-rl/spokenstorycloze 2023
6 TASLM 1B (token) 59.068.070.561.0
8 TASLM 1B (embedding) 57.567.075.550.0
9 Spirit-LM (base) 54.569.567.053.555.554.548.051.5 Spirit LM: Interleaved Spoken and Written Language Model facebookresearch/spiritlm 2024
10 pGSLM 40.583.088.557.066.053.555.553.5 Text-Free Prosody-Aware Generative Spoken Language Modeling pytorch/fairseq 2021
1–10 / 10 페이지당 10 20 50 100