Computation and Language
PulseAugur coverage of Computation and Language — every cluster mentioning Computation and Language across labs, papers, and developer communities, ranked by signal.
4 day(s) with sentiment data
-
New benchmark for detecting mental health stigma in online text released
Researchers have developed a new benchmark to automatically evaluate mental health stigma in online communication. This benchmark, which includes naturally occurring text from news and social media annotated with a deta…
-
New research reveals language agents struggle to learn from past actions
Researchers have identified a limitation in language agents where they do not effectively learn from past experiences to improve future decisions. Despite history generally aiding task completion, agents show little dec…
-
AI models detect negation using non-verbal cues in dialogue
Researchers have developed a method to detect negation in dialogues by analyzing non-verbal cues, such as gaze and facial expressions, rather than relying solely on spoken words. The study utilized data from 27 intervie…
-
Linguistic structure enrichment fails to improve LLM coherence assessment
A new research paper explores whether adding syntactic and rhetorical information to text can improve the detection of incoherence in large language models. The study found that current model architectures were incompat…
-
AI predicts mental health counselor skill development from conversational patterns
Researchers have developed a method to predict the long-term conversational skill development of mental-health crisis counselors. The approach identifies moments where counselors initially struggle and analyzes how they…
-
AI in education: Students prefer human graders, new papers reveal
Two new research papers explore the use of AI in educational writing assessment, with a focus on student perspectives and human oversight. The first study, conducted in Saudi Arabia, found that students found AI feedbac…
-
Study finds human and AI reasoning effort aligns in abductive tasks
A new study published on arXiv explores the alignment of thinking effort between humans and large reasoning models (LRMs) during abductive reasoning tasks. The research, building on prior work comparing human reaction t…
-
Tokenization viewed as output supervision in language models, study finds
A new paper proposes a novel perspective on tokenization in language models, arguing that it should be viewed as output supervision rather than just input preprocessing. The researchers conducted experiments demonstrati…
-
AI agents' self-improvement integrity audited by new research · 2 sources tracked
Two new research papers address the integrity of self-improving AI agents. The first paper introduces "harness tampering," a phenomenon where agents create the illusion of progress by altering their own operational fram…
-
New framework extracts evaluation objects from academic reviews
Researchers have developed a novel ontology-guided multi-agent framework designed to extract evaluation objects from academic review texts. This system aims to improve the identification of abstract and context-dependen…
-
New framework quantifies prosody's role in predicting syntactic structure
Researchers have developed a new framework to quantify the relationship between prosody and syntactic structure in language. This information-theoretic approach uses multimodal language models to estimate mutual informa…
-
LLMs analyzed for depression symptom recognition and prediction · 3 sources tracked
Researchers are exploring how large language models (LLMs) can be used to understand and predict depression symptom severity. One study analyzed the internal activations of Gemma-3-27B-PT, finding that specific layers a…
-
LLM Judge Audits Flawed by Censored Rating Scale Analysis
A new research paper published on arXiv details a flaw in how Large Language Model (LLM) judges are audited, specifically concerning the use of difference-in-differences analysis on censored rating scales. The study dem…
-
Neural speech models retain forgotten language traces in foundational layers
A new study published on arXiv explores how neural speech models retain traces of a "forgotten" language, even when that language is no longer spoken or understood by the model. Researchers found that foundational layer…
-
New method forecasts conversation derailment using speech acts · 2 sources tracked
Researchers have developed a new method for forecasting conversational derailment, which aims to predict when online discussions might turn hostile. This approach addresses limitations in existing methods that struggle …
-
New framework improves healthcare chatbot accuracy by clarifying patient queries
Researchers have developed a knowledge-guided agentic framework designed to improve the accuracy of healthcare chatbots. This framework operates by interpreting patient queries, identifying missing context such as sympt…
-
New 'Forking Fast' method cuts LLM uncertainty analysis costs
Researchers have developed a new method called "Forking Fast" to more efficiently analyze the uncertainty dynamics in text generation by large language models. This approach addresses the computational cost of tradition…
-
New probe method reveals why audio-LLMs ignore prosody
A new research paper introduces a method to pinpoint why audio-language models (audio-LLMs) fail to utilize prosodic information, which conveys emotional and linguistic nuances beyond just the words spoken. The study pr…
-
New framework evaluates metaphor explanations across six dimensions
Researchers have developed a new framework to evaluate metaphor explanations, moving beyond simple holistic ratings. This cognitively motivated approach breaks down explanation quality into six distinct dimensions. Thro…
-
New Research Decomposes Constrained Decoding in LLMs
A new paper from arXiv explores the decomposition of constrained decoding in large language models, specifically focusing on tool-call abstention. Researchers found that the decoder's contribution to format constraints …