# Out-of-Context Reasoning in Large Language Models

arXiv: https://arxiv.org/abs/2503.10408 · HTML: https://mlsanity.com/paper/2503.10408/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning.
- Builds on 24 of its 37 references that are in the map.
- First posted on arXiv 13 Mar 2025.

## Placed in

- LLM reasoning / [Reasoning benchmarks and evaluation](https://mlsanity.com/lr/reasoning-benchmarks-and-evaluation/)

## Builds on (most cited first, 24 of 24)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Scaling Laws for Neural Language Models](https://mlsanity.com/paper/2001.08361/) (2020)
- [Large Language Models are Zero-Shot Reasoners](https://mlsanity.com/paper/2205.11916/) (2022)
- [HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering](https://mlsanity.com/paper/1809.09600/) (2018)
- [Mistral 7B](https://mlsanity.com/paper/2310.06825/) (2023)
- [Averaging Weights Leads to Wider Optima and Better Generalization](https://mlsanity.com/paper/1803.05407/) (2018)
- [Faith and Fate: Limits of Transformers on Compositionality](https://mlsanity.com/paper/2305.18654/) (2023)
- [The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A"](https://mlsanity.com/paper/2309.12288/) (2023)
- [Logic-LM: Empowering Large Language Models with Symbolic Solvers for Faithful Logical Reasoning](https://mlsanity.com/paper/2305.12295/) (2023)
- [Large Language Models for Mathematical Reasoning: Progresses and Challenges](https://mlsanity.com/paper/2402.00157/) (2024)
- [Physics of Language Models: Part 3.1, Knowledge Storage and Extraction](https://mlsanity.com/paper/2309.14316/) (2023)
- [Mathematical Capabilities of ChatGPT](https://mlsanity.com/paper/2301.13867/) (2023)
- [More Agents Is All You Need](https://mlsanity.com/paper/2402.05120/) (2024)
- [Embers of Autoregression: Understanding Large Language Models Through the Problem They are Trained to Solve](https://mlsanity.com/paper/2309.13638/) (2023)
- [Do Large Language Models Latently Perform Multi-Hop Reasoning?](https://mlsanity.com/paper/2402.16837/) (2024)
- [Physics of Language Models: Part 3.2, Knowledge Manipulation](https://mlsanity.com/paper/2309.14402/) (2023)
- [Graph-enhanced Large Language Models in Asynchronous Plan Reasoning](https://mlsanity.com/paper/2402.02805/) (2024)
- [Exploring Transitivity in Neural NLI Models through Veridicality](https://mlsanity.com/paper/2101.10713/) (2021)
- [Rethinking Language Models as Symbolic Knowledge Graphs](https://mlsanity.com/paper/2308.13676/) (2023)
- [Beyond Memorization: The Challenge of Random Memory Access in Language Models](https://mlsanity.com/paper/2403.07805/) (2024)
- [Connecting the Dots: LLMs can Infer and Verbalize Latent Structure from Disparate Training Data](https://mlsanity.com/paper/2406.14546/) (2024)
- [Large Language Models are Limited in Out-of-Context Knowledge Reasoning](https://mlsanity.com/paper/2406.07393/) (2024)

## Built on by (most cited first, 2 of 2)

- [Tacit Coordination of Large Language Models](https://mlsanity.com/paper/2601.22184/) (2026)
- [Pro-AI Bias in Large Language Models](https://mlsanity.com/paper/2601.13749/) (2026)

## Abstract (arXiv)

We study how large language models (LLMs) reason about memorized knowledge through simple binary relations such as equality (=), inequality (<), and inclusion (subset). Unlike in-context reasoning, the axioms (e.g., a < b, b < c) are only seen during training and not provided in the task prompt (e.g., evaluating a < c). The tasks require one or more reasoning steps, and data aggregation from one or more sources, showing performance change with task complexity. We introduce a lightweight technique, out-of-context representation learning, which trains only new token embeddings on axioms and evaluates them on unseen tasks. Across reflexivity, symmetry, and transitivity tests, LLMs mostly perform statistically significant better than chance, making the correct answer extractable when testing multiple phrasing variations, but still fall short of consistent reasoning on every single query. Analysis shows that the learned embeddings are organized in structured ways, suggesting real relational understanding. Surprisingly, it also indicates that the core reasoning happens during the training, not inference.