Browse › Natural Language Processing › Discourse Parsing › RST-DT
RST-DT Benchmark (Discourse Parsing)
Over time archive 2025-07-28
The chart needs JavaScript; the table below carries every value.
Not inferred: Standard Parseval (Full), Standard Parseval (Span), Standard Parseval (Nuclearity), Standard Parseval (Relation), RST-Parseval (Full), RST-Parseval (Span), RST-Parseval (Nuclearity), RST-Parseval (Relation). Points are placed at the row's paper date; 40 of 40 rows carry one.
Results archive 2025-07-28
Archive rows end at the archive snapshot, 2025-07-28: no result published after that date is in this table. Rank is the archive's row order at that snapshot; not re-ranked here. Metric values are the archive's strings. Column headers sort the table in your browser; each row keeps its archive rank.
| Paper | Code | Ran Syntology | Report | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Bottom-up Llama 2 (70B) | 58.1 | 79.8 | 70.4 | 60.0 | – | Paper | Code | 2024 | 0 of 2 ran · 2 unverified | report | ||||
| 2 | Top-down Llama 2 (70B) | 56.0 | 78.8 | 68.7 | 57.7 | – | Paper | Code | 2024 | 0 of 2 ran · 2 unverified | report | ||||
| 3 | Bottom-up Llama 2 (13B) | 56.0 | 78.3 | 68.1 | 57.8 | – | Paper | Code | 2024 | 0 of 2 ran · 2 unverified | report | ||||
| 4 | Bottom-up Llama 2 (7B) | 55.8 | 78.2 | 67.5 | 57.6 | – | Paper | Code | 2024 | 0 of 2 ran · 2 unverified | report | ||||
| 5 | DMRST | 55.7 ± 0.3 | 78.7 ± 0.4 | 68.0 ± 0.6 | 57.3 ± 0.2 | – | Paper | Code | 2024 | linked, not harvested | report | ||||
| 6 | Top-down Llama 2 (13B) | 55.6 | 78.6 | 67.9 | 57.7 | – | Paper | Code | 2024 | 0 of 2 ran · 2 unverified | report | ||||
| 7 | Bottom-up (DeBERTa) | 55.4 ± 0.4 | 77.8 ± 0.3 | 68.0 ± 0.5 | 57.3 ± 0.2 | – | Paper | Code | 2022 | linked, not harvested | report | ||||
| 8 | Top-down (XLNet) | 54.8 | 77.8 | 67.4 | 57.0 | – | Paper | Code | 2022 | linked, not harvested | report | ||||
| 9 | Top-down (DeBERTa) | 54.4 | 78.5 | 67.9 | 56.6 | – | Paper | Code | 2022 | linked, not harvested | report | ||||
| 10 | Bottom-up (XLNet) | 54.2 | 65.9 | 56.3 | – | Paper | Code | 2022 | linked, not harvested | report | |||||
| 11 | Top-down (RoBERTa) | 53.8 | 77.3 | 66.6 | 55.8 | – | Paper | Code | 2022 | linked, not harvested | report | ||||
| 12 | Bottom-up (RoBERTa) | 53.7 | 76.1 | 66.5 | 55.4 | – | Paper | Code | 2022 | linked, not harvested | report | ||||
| 13 | Top-down Llama 2 (7B) | 53.4 | 76.3 | 65.4 | 55.2 | – | Paper | Code | 2024 | 0 of 2 ran · 2 unverified | report | ||||
| 14 | Bottom-up (SpanBERT) | 52.7 | 65.3 | 54.9 | – | Paper | Code | 2022 | linked, not harvested | report | |||||
| 15 | Top-down (SpanBERT) | 52.2 | 76.5 | 65.4 | 54.5 | – | Paper | Code | 2022 | linked, not harvested | report | ||||
| 16 | LSTM Dynamic | 50.3 | 73.1 | 62.3 | 51.5 | – | Paper | Code | 2021 | linked, not harvested | report | ||||
| 17 | End-to-end Top-down (XLNet) | 50.2 | 74.3 | 64.3 | 51.6 | 87.6 | 76.0 | 61.8 | – | Paper | Code | 2021 | linked, not harvested | report | |
| 18 | LSTM Static | 49.4 | 72.7 | 61.7 | 50.5 | – | Paper | Code | 2021 | linked, not harvested | report | ||||
| 19 | Transformer (dynamic) | 49.2 | 70.2 | 60.1 | – | Paper | Code | 2021 | linked, not harvested | report | |||||
| 20 | Transformer (static) | 49.0 | 70.6 | 59.9 | 50.6 | – | Paper | Code | 2021 | linked, not harvested | report | ||||
| 21 | End-to-end Top-down (Glove) | 46.8 | 71.1 | 59.6 | 47.7 | – | Paper | Code | 2021 | linked, not harvested | report | ||||
| 22 | Top-down (BERT) | 46.6 | 69.8 | 59.1 | 48.3 | – | Paper | Code | 2022 | linked, not harvested | report | ||||
| 23 | Bottom-up (BERT) | 46.0 | 68.3 | 57.8 | 47.8 | – | Paper | Code | 2022 | linked, not harvested | report | ||||
| 24 | Guz et al. (2020) (pretrained) | 72.94 | 61.86 | – | Paper | – | 2020 | no code linked | report | ||||||
| 25 | Guz et al. (2020) | 72.43 | 61.38 | – | Paper | – | 2020 | no code linked | report | ||||||
| 26 | Guz et al. (2020) | 72.43 | – | Paper | – | 2020 | no code linked | report | |||||||
| 27 | Top-down Span-based Parser with Silver Agreement Subtrees (ensemble) | 62.6 | 87.1 | 75.0 | 63.2 | ✓ | Paper | – | 2021 | no code linked | report | ||||
| 28 | Top-down Span-based Parser with Silver Agreement Subtrees | 61.8 | 86.8 | 74.7 | 62.5 | ✓ | Paper | – | 2021 | no code linked | report | ||||
| 29 | Transition-based Parser with Implicit Syntax Features | 59.9 | 85.5 | 73.1 | 60.2 | – | Paper | Code | 2018 | linked, not harvested | report | ||||
| 30 | HILDA Parser | 54.8 | 83.0 | 68.4 | 55.3 | – | Paper | – | 2009 | no code linked | report | ||||
| 31 | Top-down Span-based Parser | 87.0 | 74.6 | 60.0 | – | Paper | Code | 2020 | linked, not harvested | report | |||||
| 32 | Two-stage Parser | 86.0 | 72.4 | 59.7 | – | Paper | Code | 2017 | linked, not harvested | report | |||||
| 33 | Bottom-up Linear-chain CRF-based Parser | 85.7 | 71.0 | 58.2 | – | Paper | – | 2014 | no code linked | report | |||||
| 34 | Two-stage Discourse Parser with a Sliding Window | 83.84 | 68.90 | 55.87 | – | Paper | – | 2015 | no code linked | report | |||||
| 35 | Greedy Bottom-up Parser with Syntactic Features | 54.9* | 82.6* | 67.1* | 55.4* | – | Paper | – | 2015 | no code linked | report | ||||
| 36 | Re-implemented HILDA RST parser | 54.3* | 82.6* | 66.6* | 54.6* | – | Paper | – | 2016 | no code linked | report | ||||
| 37 | Discourse Parser with Hierarchical Attention | 50.6* | 82.2* | 66.5* | 51.4* | – | Paper | – | 2016 | no code linked | report | ||||
| 38 | Discourse Parsing from Linear Projection | 57.6* | 82.0* | 68.2* | 57.8* | – | Paper | Code | 2014 | linked, not harvested | report | ||||
| 39 | Transition-Based Parser Trained on Cross-Lingual Corpus | 56.0* | 81.3* | 68.1* | 56.3* | – | Paper | Code | 2017 | linked, not harvested | report | ||||
| 40 | LSTM Sequential Discourse Parser (Braud et al., 2016) | 47.5* | 79.7* | 63.6* | 47.7* | – | Paper | Code | 2016 | linked, not harvested | report |
All 40 rows shown. 40 link to a paper page on this site; 2 are marked as using additional training data in the archive. No GitHub stars are tracked; "Code" is the first repository the archive lists for the row. The archive carries no row tags, review links or community-submitted rows for this table; none are shown. archive 2025-07-28
Syntology Ran reads "N of M ran · U unverified": of the M code samples Syntology harvested from repositories linked to that row's paper (joined by arXiv id), N executed on a synthesized input and the other U = M−N are unverified (harvested, no recorded run). It counts code from repositories linked to that row's paper, not this result: the row's number was not reproduced and nothing here is a correctness claim. The other cell texts mean no graph line for the row: "linked, not harvested" (the archive links code, Syntology has not harvested it), "no code linked" (no code link in the archive), "not matched" (the row's paper URL matched no paper on this site). 6 rows have a graph line, from 1 distinct papers; 0 rows (0 papers) have at least one sample that ran. Counting each paper once: Syntology ran 0 of 2 samples; 2 unverified. Separately, 2 of those 2 are pointer-only (licence): the site points at that code rather than redistributing it, a licence property recorded for ran and unverified samples alike; each cell's tooltip carries the row's own pointer-only count. Read from the graph 2026-09-25. Per-sample status is on the paper page.
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections