Search Results - "Bhattacharya, Sunit"

Understanding the role of FFNs in driving multilingual behaviour in LLMs by Bhattacharya, Sunit, Bojar, Ondřej

“…Multilingualism in Large Language Models (LLMs) is an yet under-explored area. In this paper, we conduct an in-depth analysis of the multilingual capabilities…”
Get full text

Unveiling Multilinguality in Transformer Models: Exploring Language Specificity in Feed-Forward Networks by Bhattacharya, Sunit, Bojar, Ondrej

“…Recent research suggests that the feed-forward module within Transformers can be viewed as a collection of key-value memories, where the keys learn to capture…”
Get full text

Multimodal Shannon Game with Images by Zouhar, Vilém, Bhattacharya, Sunit, Bojar, Ondřej

“…The Shannon game has long been used as a thought experiment in linguistics and NLP, asking participants to guess the next letter in a sentence based on its…”
Get full text

Sentence Ambiguity, Grammaticality and Complexity Probes by Bhattacharya, Sunit, Zouhar, Vilém, Bojar, Ondřej

“…It is unclear whether, how and where large pre-trained language models capture subtle linguistic traits like ambiguity, grammaticality and sentence complexity…”
Get full text

Team \'UFAL at CMCL 2022 Shared Task: Figuring out the correct recipe for predicting Eye-Tracking features using Pretrained Language Models by Bhattacharya, Sunit, Kumar, Rishu, Bojar, Ondrej

“…Eye-Tracking data is a very useful source of information to study cognition and especially language comprehension in humans. In this paper, we describe our…”
Get full text

EMMT: A simultaneous eye-tracking, 4-electrode EEG and audio corpus for multi-modal reading and translation scenarios by Bhattacharya, Sunit, Kloudová, Věra, Zouhar, Vilém, Bojar, Ondřej

“…We present the Eyetracked Multi-Modal Translation (EMMT) corpus, a dataset containing monocular eye movement recordings, audio and 4-electrode…”
Get full text

Coreference between subjective expression and holder: A classification perspective by Das, Dipankar, Bhattacharya, Sunit

“…Instead of applying the traditional keyword spotting approach, the presence of coreference between the holders and their subjective expressions often becomes…”
Get full text