AI Paper Atlas

Research map / Large language models / AI and society

Can LLMs Move Beyond Short Exchanges to Realistic Therapy Conversations?

Zhengqing Yuan, Liang Wu, Jian Xu, Zheyuan Zhang, Kaiwen Shi, Weixiang Sun, Lichao Sun, Yanfang Ye

ICLR 2026 · Large language models › AI and society · approach: Critiques of LLM claims and judging

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
rephrase public exam items by clinicians18 LLMs · CareBench-CBT knowledge QAaccuracy: 92.31 -> 71.17Hurt
add Hierarchical Therapy Memory (HTM) (ablation)GPT-5 · CareBench-CBT multi-turnwhole-session score: 4.10 -> 4.45Improved
add HTM (ablation)LLaMA-3.3-70B · CareBench-CBT multi-turnturn score (model-based history): 1.47 -> 1.85Improved
add HTM (ablation)Qwen-2.5-7B · CareBench-CBT multi-turnturn score (model-based history): 1.93 -> 1.97No clear effect
add HTM (ablation)GPT-5 · CareBench-CBT multi-turnnormalized latency: 1.12x -> 0.65xImproved
use human counselor history instead of model-generated history (ablation)18 LLMs · CareBench-CBT multi-turnturn score: average gap -0.01No clear effect
use human-based history instead of model history (ablation)GPT-oss-120B · CareBench-CBT multi-turnturn score: -0.40Hurt
scale to large open-weight model on vignette classificationGPT-oss-120B · CareBench-CBT case vignettesaccuracy: 41.7%No clear effect
add explicit length constraints or brevity promptsLLMs · CareBench-CBT multi-turncounseling qualityHurt

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