AI Paper Atlas

Research map / Large language models / Alignment and preferences

Evolution and compression in LLMs: on the emergence of human-aligned categorization

Nathaniel Imel, Noga Zaslavsky

ICLR 2026 · Large language models › Alignment and preferences · approach: Steering and correcting LLM behavior · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
scale up and instruction-tune the LLMLLMs on English color naming · English color naming (WCS, 39 models)English-alignment and IB-efficiencyImproved
instruction-tune a very large model (Llama 3.3 70B)Llama 3.3 70B · English color namingIB-efficiency / English alignmentNo clear effect
present colors as images instead of sRGB textlarger multimodal LLMs · English color namingalignment and efficiencyNo clear effect
present colors as images instead of sRGB textsmaller multimodal LLMs · English color namingalignment and efficiencyImproved
present colors as CIELAB triples instead of sRGBtop four LLMs · English color namingEnglish alignmentHurt
run iterated in-context language learning over generationsGemini 2.0, Gemma 3 27B, Qwen 2.5 32B, Llama 3.3 70B · IICLL color namingIB-efficiencyImproved
use Gemini 2.0 instead of other LLMs in IICLLIICLL · IICLL color namingcomplexity range reachedImproved
use conversation history window of 10 instead of 0 (ablation)IICLL · IICLL chainsdegenerate systemsImproved
increase history window from 10 to 20 or 50 (ablation)IICLL · IICLL chainsresultsNo clear effect
use small instruction-tuned models in IICLLGemma 3 4b, Llama 3.2 3B, Qwen 2.5 3B, OLMo 2 7B · IICLL color namingnon-degenerate systems: 19 of 20 chains degenerateHurt
instruction-tuning stage in trainingOlmo 2 32B · English color namingEnglish-alignmentImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy