REVIEW 14 cited by
Aya 23: Open Weight Releases to Further Multilingual Progress
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Aya 23: Open Weight Releases to Further Multilingual Progress
read the original abstract
This technical report introduces Aya 23, a family of multilingual language models. Aya 23 builds on the recent release of the Aya model (\"Ust\"un et al., 2024), focusing on pairing a highly performant pre-trained model with the recently released Aya collection (Singh et al., 2024). The result is a powerful multilingual large language model serving 23 languages, expanding state-of-art language modeling capabilities to approximately half of the world's population. The Aya model covered 101 languages whereas Aya 23 is an experiment in depth vs breadth, exploring the impact of allocating more capacity to fewer languages that are included during pre-training. Aya 23 outperforms both previous massively multilingual models like Aya 101 for the languages it covers, as well as widely used models like Gemma, Mistral and Mixtral on an extensive range of discriminative and generative tasks. We release the open weights for both the 8B and 35B models as part of our continued commitment for expanding access to multilingual progress.
Forward citations
Cited by 14 Pith papers
-
CrossHallu: Do Hallucination Signals Generalize Across Languages and Domains in Large Language Model's Internals?
Hallucination signals from LLM internals transfer across English–Arabic and Arabic domains for most models, depending on class separability and feature-space language alignment.
-
When Similar Means Different: Evaluating LLMs on Arabic--Hebrew Cognates
LLMs achieve high accuracy on true Arabic-Hebrew cognates but drop sharply on false friends and loanwords due to surface-form reliance, with only modest gains from sentence context.
-
SomaliBench Eval: Measuring English-to-Somali Refusal Gaps in Open-Weight Language Models
SomaliBench finds large English-to-Somali refusal gaps (0.38 to 0.90) across Llama-3.1-8B, Gemma-2-9B, Qwen-2.5-7B, and Aya-23-8B, with many Somali responses being unclear rather than compliant.
-
Vocab Diet: Reshaping the Vocabulary of LLMs via Vector Arithmetic
LLMs can compose surface-form tokens from base embeddings plus learned transformation vectors, freeing 10-40% of vocabulary slots while expanding coverage and preserving downstream performance across five languages.
-
ChiKhaPo: A Large-Scale Multilingual Benchmark for Evaluating Lexical Comprehension and Generation in Large Language Models
ChiKhaPo is an 8-subtask benchmark that measures word-level comprehension and generation in 2,700+ languages and shows state-of-the-art models perform poorly on low-resource languages.
-
Pancasila-Dilemmas: Evaluating Large Language Models on Indonesian Human Value Dilemmas Grounded in Pancasila
A 1,834-question Indonesian dilemma benchmark shows leading LLMs match Indonesian human choices only about half the time and struggle most on religion and unity.
-
Sentence-Level Contextual Entrainment in Large Language Models
Sentence-level contextual entrainment exists across LLMs, weakens with scale, and is localized to 2-4% of attention heads whose deactivation removes the effect without performance loss.
-
Lost in the Tower of Babel: The Adverse Effects of Incidental Multilingualism in LLMs
Frontier LLMs' self-declared language support is unstable and over-optimistic, verified behavior is task-dependent, and language mismatch alone degrades collaborative agent performance.
-
Linguistic Nepotism: Trading-off Quality for Language Preference in Multilingual RAG
In multilingual retrieval-augmented generation, models cite English evidence more accurately than translated evidence, and this language preference can outweigh document relevance.
-
Speaking in Self-Assessing Tongues: On the Verbalized Confidence of LLMs in Machine Translation
Empirical study finds verbalized per-token confidence methods in LLMs for MT perform similarly to internal signals on error detection and calibration but show little correlation.
-
Lost in the Tower of Babel: The Adverse Effects of Incidental Multilingualism in LLMs
Incidental multilingualism from uneven web training makes LLMs unequal, brittle, and opaque across languages.
-
Patching LLM Like Software: A Lightweight Method for Improving Safety Policy in Large Language Models
Prepending a compact learnable prefix to LLMs produces safety gains comparable to next-generation aligned models while preserving fluency and adding negligible parameters.
-
Optimizing Korean-Centric LLMs via Token Pruning
Token pruning of non-Korean vocabulary in LLMs improves generation stability and often boosts machine translation on Korean tasks while cutting vocabulary size substantially.
-
Multilingual Vision-Language Models, A Survey
The survey identifies a key tension in multilingual vision-language models between language neutrality via contrastive learning and cultural awareness via diverse data, with most benchmarks relying on translation-base...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.