Computation and Language · arXiv:2409.06173 v3 · submitted 10 Sep 2024
Larger Language Models Don't Care How You Think: Why Chain-of-Thought Prompting Fails in Subjective Tasks
The paper says
“We find that, surprisingly, CoT indeed suffers from the same posterior collapse as ICL for larger language models.”
from the abstract · picked out as the main claim by a classifier
Reviews
Reviews by people
No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.
The paper
Abstract
In-Context Learning (ICL) in Large Language Models (LLM) has emerged as the dominant technique for performing natural language tasks, as it does not require updating the model parameters with gradient-based methods. ICL promises to "adapt" the LLM to perform the present task at a competitive or state-of-the-art level at a fraction of the computational cost. ICL can be augmented by incorporating the reasoning process to arrive at the final label explicitly in the prompt, a technique called Chain-of-Thought (CoT) prompting. However, recent work has found that ICL relies mostly on the retrieval of task priors and less so on "learning" to perform tasks, especially for complex subjective domains like emotion and morality, where priors ossify posterior predictions. In this work, we examine whether "enabling" reasoning also creates the same behavior in LLMs, wherein the format of CoT retrieves reasoning priors that remain relatively unchanged despite the evidence in the prompt. We find that, surprisingly, CoT indeed suffers from the same posterior collapse as ICL for larger language models. Code is avalaible at https://github.com/gchochla/cot-priors.
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.
abstract compiled · 2026-09-23 · no language-model review
How the main claim is stated: asserted 1.00
What the abstract itself states, as probabilities: claims first 0.05 · states limitation 0.34 · human participants 0.07 · controlled comparison 0.38 · randomized assignment 0.01 · states quantitative result 0.04 · compares with prior results 0.89 · mentions code or data availability 0.99
The abstract, sentence by sentence
This table labels the abstract only (arXiv listing abstract v3). The number is the probability the classifier gave the label.
| Sentence | Role | The authors' own work | States the main conclusion | Strength |
|---|---|---|---|---|
| In-Context Learning (ICL) in Large Language Models (LLM) has emerged as the dominant technique for performing natural language tasks, as it does not require updating the model parameters with gradient-based methods. | background 1.00 | 0.04 | 0.02 | asserted 1.00 |
| ICL promises to "adapt" the LLM to perform the present task at a competitive or state-of-the-art level at a fraction of the computational cost. | background 1.00 | 0.05 | 0.02 | hedged possibility 0.51 |
| ICL can be augmented by incorporating the reasoning process to arrive at the final label explicitly in the prompt, a technique called Chain-of-Thought (CoT) prompting. | background 0.99 | 0.04 | 0.02 | asserted 0.89 |
| However, recent work has found that ICL relies mostly on the retrieval of task priors and less so on "learning" to perform tasks, especially for complex subjective domains like emotion and morality, where priors ossify posterior predictions. | background 0.94 | 0.05 | 0.07 | asserted 0.97 |
| In this work, we examine whether "enabling" reasoning also creates the same behavior in LLMs, wherein the format of CoT retrieves reasoning priors that remain relatively unchanged despite the evidence in the prompt. | contribution 0.88 | 0.25 | 0.05 | hedged possibility 0.44 |
| We find that, surprisingly, CoT indeed suffers from the same posterior collapse as ICL for larger language models. | result 1.00 | 0.96 | 0.91 | asserted 1.00 |
| Code is avalaible at https://github.com/gchochla/cot-priors. | contribution 0.65 | 0.21 | 0.01 | asserted 0.64 |
What kind of paper the classifier reports this is: empirical study 1.00 · evidence from observation or experiment 0.94
What was examined, and what was not
Sources read
- arXiv listing abstract v3
Not examined
- full text
- figures and tables
- references and citing sentences
- relations to other papers
Levels run
- P abstract
Not run
- S sections
- T body sentences
- C citations
- R cross-paper relations
How the labels were made
Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-09-23: 1 request, 5,685 input tokens.
Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON
Cited by 2 Pith papers
Reviewed papers on Pith that cite this one. The full citation record
How the literature uses it
1 arXiv paper cite this in 2 sentences
Papers citing it · 1 sentence without a confident label
Cites it as support 1 paper
-
However, such models are still inadequate at understanding intricate social reasoning and human contexts [ref] .
CaRGo-T: Causal Reasoning Graph-of-Thought improves Multimodal Humor Comprehension · Related Work · 0.98
What it cites · 8 sentences without a confident label
Uses the method or tool of 1 paper
-
We closely follow the methodology and notation in [ref] .
The Strong Pull of Prior Knowledge in Large Language Models and Its Impact on Emotion Recognition · Methodology · 1.00
Uses data or values from 1 paper
-
MFRC [ref] : Multilabel moral foundation corpus with annotations for six moral foundations: care, equality, proportionality, loyalty, authority, and purity.
2208.05545 · Datasets · 0.98
Cites as background 1 paper
-
Large Language Models (LLMs) [ref] have come to dominate language processing tasks as tools that are reliable, affordable, and scalable in many disciplines [ref] .
The Llama 3 Herd of Models · Introduction · 0.99
Challenges and discussion
Know something about this paper? Leave a comment.
Come back by email
A short list of the week's newly reviewed papers. Nothing is sent until you confirm from your inbox, and every email has a link that stops it. No account.
No comments yet.
Sign in to comment