mlsanityData to 7 Oct 2026

Probing Latent Colombian Identity Inferences in Qwen2.5-7B with Natural Language Autoencoders

arXiv 2607.21774 ↗First posted 23 Jul 2026cs.CL cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Large language models may infer demographic attributes from subtle linguistic cues even when those attributes are not explicitly stated. This pilot study examines whether Qwen2.5-7B-Instruct internally represents Colombian identity, socioeconomic status, or stereotype-related information when processing Colombian-Spanish and English prompts. We use Natural Language Autoencoders (NLA) to verbalize residual-stream activations from layer 20 across four positional quartiles per prompt. Our dataset contains 30 prompts arranged as 15 matched Spanish-English pairs, spanning explicit Colombian cues, implicit Colombian cues, and neutral controls. We report descriptive rates and qualitative evidence rather than statistically powered effects, focusing on whether latent nationality or stereotype representations appear before they are verbalized in the model output. This work connects activation-level interpretability with bias evaluation for underrepresented Spanish varieties.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 23 of its 28 references that are in the map. First posted on arXiv 23 Jul 2026.

Builds on

23 of 28 references.

References · 23

  1. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,359
  2. 2023Eliciting Latent Predictions from Transformers with the Tuned Lenscited by 579
  3. 2024Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Modelscited by 225
  4. 2020Detecting Emergent Intersectional Biases: Contextualized Word Embeddings Contain a Distribution of Human-like Biasescited by 167
  5. 2019Reducing Sentiment Bias in Language Models via Counterfactual Evaluationcited by 148
  6. 2019Assessing Social and Intersectional Biases in Contextualized Word Representationscited by 134
  7. 2024Unintended Impacts of LLM Alignment on Global Representationcited by 66
  8. 2024SelfIE: Self-Interpretation of Large Language Model Embeddingscited by 63
  9. 2024Linguistic Bias in ChatGPT: Language Models Reinforce Dialect Discriminationcited by 55
  10. 2024LatentQA: Teaching LLMs to Decode Activations Into Natural Languagecited by 49
  11. 2025Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainerscited by 41
  12. 2023Quantifying the Dialect Gap and its Correlates Across Languagescited by 33
11 more
  1. 2025Reading Between the Prompts: How Stereotypes Shape LLM's Implicit Personalizationcited by 24
  2. 2025Rejected Dialects: Biases Against African American Language in Reward Modelscited by 20
  3. 2023What Do Llamas Really Think? Revealing Preference Biases in Language Model Representationscited by 10
  4. 2022SocioProbe: What, When, and Where Language Models Learn about Sociodemographicscited by 7
  5. 2024Native Design Bias: Studying the Impact of English Nativeness on Language Model Performancecited by 5
  6. 2025It's the same but not the same: Do LLMs distinguish Spanish varieties?cited by 3
  7. 2025Linear socio-demographic representations emerge in Large Language Models from indirect cuescited by 2
  8. 2026How Does "English (US)" Become the Default? Triangulating Structural Bias Towards American English Across the LLM Pipelinecited by 2
  9. 2025SESGO: Spanish Evaluation of Stereotypical Generative Outputscited by 1
  10. 2026Tracing the Latent Threads: A Mechanistic Study of How LLMs Represent and Operationalize Race and Ethnicity Cuescited by 1
  11. 2026Digital Linguistic Bias in Spanish: Evidence from Lexical Variation in LLMscited by 1

Built on by

Not cited yet.