Corrections: 2025 January (#4399)

* Paper Revision: 2024.clasp-1.7, closes #4370.

* Paper Revision 2024.emnlp-main.1163, closes #4394.

* Paper Revision{2024.naacl-long.268}, closes #4325.

* Paper Revision 2024.lrec-main.561, closes #4310.

* Paper Revision 2024.kallm-1.7, closes #4308.

* Paper Revision{2023.findings-acl.275}, closes #4190.

* Paper Revision: 2023.findings-emnlp.501, closes #4189.

* Paper Revision{2024.emnlp-main.468}, closes #4179.

* Paper Revision: {2024.emnlp-main.304}, closes #4166.

* Paper Revision{2024.emnlp-main.572}, closes #4165.

* Paper Revision{2024.emnlp-main.1248}, closes #4163.

* Paper Revision{2024.emnlp-main.1238}, closes #4162.

* Paper Revision{2024.findings-emnlp.144}, closes #4159.

* Paper Revision{2023.arabicnlp-1.15}, closes #4157.

* Paper Revision: 2024.lrec-main.959, closes #4152.

* Paper Revision: 2022.coling-1.525, closes #4497.

* Paper Revision: {2024.findings-naacl.159}, closes #4156.

* Add William Soto Martinez (closes #4043)

* Automatically write sorted name variants file

---------

Co-authored-by: Matt Post <[email protected]>

bin/clean_name_variants.py

@@ -35,4 +35,5 @@
 
 newvariants.sort(key=lambda v: (v["canonical"]["last"], v["canonical"]["first"]))
 
-print(yaml.dump(newvariants, allow_unicode=True, default_flow_style=None))
+with open(os.path.join(datadir, "yaml", "name_variants.yaml"), "wt") as f:
+    print(yaml.dump(newvariants, allow_unicode=True, default_flow_style=None), file=f)

data/xml/2022.coling.xml

@@ -6926,13 +6926,14 @@
       <author><first>Ryan</first><last>Cotterell</last></author>
       <pages>6007–6018</pages>
       <abstract>Recombining known primitive concepts into larger novel combinations is a quintessentially human cognitive capability. Whether large neural models in NLP acquire this ability while learning from data is an open question. In this paper, we look at this problem from the perspective of formal languages. We use deterministic finite-state transducers to make an unbounded number of datasets with controllable properties governing compositionality. By randomly sampling over many transducers, we explore which of their properties (number of states, alphabet size, number of transitions etc.) contribute to learnability of a compositional relation by a neural network. In general, we find that the models either learn the relations completely or not at all. The key is transition coverage, setting a soft learnability limit at 400 examples per transition.</abstract>
-      <url hash="602a0e58">2022.coling-1.525</url>
+      <url hash="b1d5e54f">2022.coling-1.525</url>
       <bibkey>valvoda-etal-2022-benchmarking</bibkey>
       <revision id="1" href="2022.coling-1.525v1" hash="6754108c"/>
       <revision id="2" href="2022.coling-1.525v2" hash="602a0e58" date="2023-07-31">Various corrections.</revision>
       <pwccode url="https://github.com/valvoda/neuraltransducer" additional="false">valvoda/neuraltransducer</pwccode>
       <pwcdataset url="https://paperswithcode.com/dataset/gscan">GSCAN</pwcdataset>
       <pwcdataset url="https://paperswithcode.com/dataset/scan">SCAN</pwcdataset>
+      <revision id="3" href="2022.coling-1.525v3" hash="b1d5e54f" date="2025-01-28">Minor updates.</revision>
     </paper>
     <paper id="526">
       <title>Source-summary Entity Aggregation in Abstractive Summarization</title>

data/xml/2023.arabicnlp.xml

@@ -208,9 +208,11 @@
       <author><first>Nizar</first><last>Habash</last></author>
       <pages>170-180</pages>
       <abstract>We present CamelParser2.0, an open-source Python-based Arabic dependency parser targeting two popular Arabic dependency formalisms, the Columbia Arabic Treebank (CATiB), and Universal Dependencies (UD). The CamelParser2.0 pipeline handles the processing of raw text and produces tokenization, part-of-speech and rich morphological features. As part of developing CamelParser2.0, we explore many system design hyper-parameters, such as parsing model architecture and pretrained language model selection, achieving new state-of-the-art performance across diverse Arabic genres under gold and predicted tokenization settings.</abstract>
-      <url hash="f6e5ed81">2023.arabicnlp-1.15</url>
+      <url hash="cd7ace87">2023.arabicnlp-1.15</url>
       <bibkey>elshabrawy-etal-2023-camelparser2</bibkey>
       <doi>10.18653/v1/2023.arabicnlp-1.15</doi>
+      <revision id="1" href="2023.arabicnlp-1.15v1" hash="f6e5ed81"/>
+      <revision id="2" href="2023.arabicnlp-1.15v2" hash="cd7ace87" date="2025-01-14">Minor updates.</revision>
     </paper>
     <paper id="16">
       <title><fixed-case>GARI</fixed-case>: Graph Attention for Relative Isomorphism of <fixed-case>A</fixed-case>rabic Word Embeddings</title>

data/xml/2023.findings.xml

@@ -6358,9 +6358,11 @@
       <author><first>Pascale</first><last>Fung</last><affiliation>Hong Kong University of Science and Technology</affiliation></author>
       <pages>4504-4522</pages>
       <abstract>Dialogue systems can leverage large pre-trained language models and knowledge to generate fluent and informative responses. However, these models are still prone to produce hallucinated responses not supported by the input source, which greatly hinders their application. The heterogeneity between external knowledge and dialogue context challenges representation learning and source integration, which further contributes to unfaithfulness. To handle this challenge and generate more faithful responses, this paper presents RHO (ρ) utilizing the representations of linked entities and relation predicates from a knowledge graph (KG). We propose (1) local knowledge grounding to combine textual embeddings with the corresponding KG embeddings; and (2) global knowledge grounding to equip RHO with multi-hop reasoning abilities via the attention mechanism. In addition, we devise a response re-ranking technique based on walks over KG sub-graphs for better conversational reasoning. Experimental results on OpenDialKG (Moon et al., 2019) show that our approach significantly outperforms state-of-the-art methods on both automatic and human evaluation by a large margin, especially in hallucination reduction (17.54% in FeQA (Durmus et al., 2020)).</abstract>
-      <url hash="30f72dcd">2023.findings-acl.275</url>
+      <url hash="8f36fcb1">2023.findings-acl.275</url>
       <bibkey>ji-etal-2023-rho</bibkey>
       <doi>10.18653/v1/2023.findings-acl.275</doi>
+      <revision id="1" href="2023.findings-acl.275v1" hash="30f72dcd"/>
+      <revision id="2" href="2023.findings-acl.275v2" hash="8f36fcb1" date="2025-01-12">Minor updates.</revision>
     </paper>
     <paper id="276">
       <title>Transformer Language Models Handle Word Frequency in Prediction Head</title>
@@ -21158,11 +21160,12 @@
       <author><first>Kam-Fai</first><last>Wong</last></author>
       <pages>7468-7479</pages>
       <abstract>Knowledge probing is a task to assess the knowledge encoded within pre-trained language models (PLMs) by having the PLM complete prompts such as “Italy is located in __,”. The model’s prediction precision serves as a lower bound for the amount of knowledge it contains. Subsequent works explore training a series of vectors as prompts to guide PLMs towards more accurate predictions. However, these methods compromise the readability of the prompts. We cannot directly understand these prompts from their literal meaning, making it difficult to verify whether they are correct. Consequently, the credibility of probing results derived from these prompts is diminished. To address the issue, we propose a novel method called ReadPrompt, which aims to identify meaningful sentences to serve as prompts. Experiments show that ReadPrompt achieves state-of-the-art performance on the current knowledge probing benchmark. Moreover, since the prompt is readable, we discovered a misalignment between constructed prompts and knowledge, which is also present in current prompting methods verified by an attack experiment. We claim that the probing outcomes of the current prompting methods are unreliable that overestimate the knowledge contained within PLMs.</abstract>
-      <url hash="2063ba71">2023.findings-emnlp.501</url>
+      <url hash="f9e4eba0">2023.findings-emnlp.501</url>
       <bibkey>wang-etal-2023-readprompt</bibkey>
       <doi>10.18653/v1/2023.findings-emnlp.501</doi>
       <revision id="1" href="2023.findings-emnlp.501v1" hash="5f48a99b"/>
       <revision id="2" href="2023.findings-emnlp.501v2" hash="2063ba71" date="2024-02-28">Updated acknowledgment.</revision>
+      <revision id="3" href="2023.findings-emnlp.501v3" hash="f9e4eba0" date="2025-01-12">Revised acknowledgements.</revision>
     </paper>
     <paper id="502">
       <title>Coherent Entity Disambiguation via Modeling Topic and Categorical Dependency</title>

data/xml/2024.clasp.xml

@@ -86,8 +86,10 @@
       <author><first>Sina</first><last>Zarrieß</last></author>
       <pages>39–55</pages>
       <abstract>Syntactic learning curves in LMs are usually reported as relatively stable and power law-shaped. By analyzing the learning curves of different LMs on various syntactic phenomena using both small self-trained llama models and larger pre-trained pythia models, we show that while many phenomena do follow typical power law curves, others exhibit S-shaped, U-shaped, or erratic patterns. Certain syntactic paradigms remain challenging even for large models, resulting in persistent preference for ungrammatical sentences. Most phenomena show similar curves for their paradigms, but the existence of diverging patterns and oscillations indicates that average curves mask important developments, underscoring the need for more detailed analyses of individual learning trajectories.</abstract>
-      <url hash="def43411">2024.clasp-1.7</url>
+      <url hash="c8f43866">2024.clasp-1.7</url>
       <bibkey>bunzeck-zarriess-2024-fifty</bibkey>
+      <revision id="1" href="2024.clasp-1.7v1" hash="def43411"/>
+      <revision id="2" href="2024.clasp-1.7v2" hash="c8f43866" date="2025-01-12">This revision corrects an axis-scaling error for the three smaller Pythia models in Figure 2 and adjusts the description of the Figure and its results accordingly.</revision>
     </paper>
     <paper id="8">
       <title>Not Just Semantics: Word Meaning Negotiation in Social Media and Spoken Interaction</title>