statusnotesummaryupdated
draftEvolutionary Strategies lead to Catastrophic Forgetting in LLMsWarns that ES post-training can cause catastrophic forgetting through large dense updates.2026-07-31
draftA Little Rank Goes a Long Way: Random Scaffolds with LoRA Adapters Are All You NeedStudies how much task information low-rank adapters can carry over frozen random backbones.2026-07-31
draftMatching Accuracy, Different Geometry: Evolution Strategies vs GRPO in LLM Post-TrainingShows ES and GRPO can match accuracy while reaching different update geometries.2026-07-31
draftLoRA: Low-Rank Adaptation of Large Language ModelsIntroduces LoRA, the low-rank adapter baseline for parameter-efficient LLM adaptation.2026-07-31
draftDoRA: Weight-Decomposed Low-Rank AdaptationDoRA separates pretrained weight magnitude from direction and applies LoRA-style updates to direction.2026-07-31
draftEvolution Strategies at Scale: LLM Fine-Tuning Beyond Reinforcement LearningArgues that full-parameter ES can scale to billion-parameter LLM fine-tuning.2026-07-31
draftEvolution Strategies at the HyperscaleEGGROLL structures ES perturbations as low-rank matrices to improve large-scale black-box optimisation.2026-07-31
draftOn the Relationship Between the OpenAI Evolution Strategy and Stochastic Gradient DescentCompares OpenAI ES with SGD through gradient-correlation and proxy experiments.2026-07-31
draftEmpirical sketch proofsEmpirical evidence ladder from EGGROLL port validation to matrix and small-LM synthetic objectives.2026-08-01
draftExperiment: llm_tests_jean_zayMerged LLM smoke tests and pretrained small-LM logprob sweep for EGGROLL vs decomposed EGGROLL.2026-08-01
draftExperiment: matrix_tests_jean_zayMerged matrix smoke tests and robustness sweep for plain EGGROLL vs decomposed EGGROLL.2026-08-01
draftStarting pointDraft framing and baseline equations for DoRA-style decomposed ES on long-context reward adaptation.2026-08-01

In one paragraph

This project asks whether evolution strategies can become a credible tool for adapting LLM agents from long-context, reward-only feedback by searching in a better low-rank parameterisation. The working idea combines EGGROLL-style low-rank ES perturbations with a DoRA-style weight decomposition, where magnitude and direction are adapted separately. The implementation has now passed matrix and small-LM synthetic gates; the next decisive step is to replace answer-token likelihood with scalar exact-answer reward.

Why it matters

Long-context agent training often produces feedback that is sparse, delayed, noisy, or hard to differentiate through. Gradient-free methods such as ES are attractive in that setting, but naive search in high-dimensional adapter coordinates can be inefficient or unstable. If magnitude/direction decomposition improves the effective search geometry, the project could give a simple empirical and theoretical reason to use decomposed adapters for reward-driven LLM adaptation.

Approach

The baseline is plain LoRA-space EGGROLL: freeze the base model and perturb low-rank adapter parameters. The proposed method keeps the low-rank ES estimator but searches in a DoRA-like parameterisation, separating output-channel magnitude from directional low-rank updates. The empirical ladder is deliberately conservative: matrix plumbing, matrix robustness, causal-LM plumbing, pretrained small-LM logprob optimisation, then tiny real-reward tasks before any long-context agent benchmark.

Current findings

The matrix tests are the strongest positive signal so far. Plain EGGROLL learned the controlled low-rank matrix objective, and full decomposed EGGROLL was more reliable in the seed/rank sweep; direction-only and magnitude-only ablations failed. The LLM tests show that both methods can patch causal-LM modules, stay finite, and improve synthetic NLL objectives on random and pretrained small models. The latest pretrained sweep was stable and mildly favorable to decomposed EGGROLL, but it is still generated arithmetic logprob optimisation, not reward-only behavioural learning.

See Empirical sketch proofs, matrix tests, and LLM tests.

5 items under this folder.