The Illusion of Latent Generalization: Bi-directionality and the Reversal Curse
Julian Coda-Forno, Jane X. Wang, Arslan Chaudhry · 2026
Read The Illusion of Latent Generalization: Bi-directionality and the Reversal Curse by Julian Coda-Forno, Jane X. Wang, Arslan Chaudhry (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-04943.txt.
Description
Coda-Forno, Wang, and Chaudhry examine the reversal curse, where models trained on a fact in one direction fail to retrieve it reversed. Bidirectional supervision, such as bidirectional attention or masking-based reconstruction, can ease the curse. The authors extend the test to masked language modeling and compare it with decoder-only masking across four benchmarks. They find reversal accuracy needs training signal that makes the source entity a prediction target. Success, they argue, shows little sign of a direction-free representation of a fact.
Catalog record
| CREATOR | Julian Coda-Forno, Jane X. Wang, Arslan Chaudhry |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 26555 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.04943