Cross-base transfer: B10 transfers most robustly

B10 fine-tuning yields smallest OOD drops when transferring to other bases

Source paper

extracted_from

(2025) · Edward Yi Chang · Kaya, Zeyneb N. · Ethan Chang

claim

community

Few-shot anchoring & latent structure
members_of
How minimal examples disambiguate and recruit latent arithmetic/reasoning interpretations in LLMs
Model base robustness and transfer learning asymmetries
members_of
Investigates how neural network bases (B10 vs B9) differ in fine-tuning stability and cross-base transfer, with density as a key predictor of robustness.
Cross-base fine-tuning transfer asymmetry
members_of
Empirical finding that base-10 representations transfer more robustly out-of-distribution than base-9 after fine-tuning.

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

Cross-base fine-tuning yields asymmetric transfer: B10 transfers most robustly, B9 leastfinding0.865
In-base gains accompanied by uneven OOD drops; higher-density priors more robust.
Cross-base transferconcept0.862
Transfer of arithmetic rules from one numeral base to another after SFT.
cross-base interferenceconcept0.760
Asymmetric transfer after fine-tuning: high-density bases (B10) are more robust.
Higher-density priors (B10) are more robust to cross-base OOD drops than lower-density ones (B9) after fine-tuningfinding0.746
E2 asymmetric transfer finding consistent with UCCT's mismatch-driven OOD fragility
B10 final accuracy 94.8 ± 1.2%finding0.726
Accuracy at k=16 shots for B10.
Base-10 additionconcept0.720
The generic addition mechanism that Llama-3.1-8B actually uses to compute sums before mapping back to cyclic concept space
Higher-density priors (B10) are more robust to fine-tuning than lower-density ones (B9).claim0.712
Interpretation of cross-base transfer asymmetry.
Stress Transferconcept0.698
Mechanism by which stress becomes mutually transferable between humans and technology, enabling symbiotic integration.