autoresearch · Issues· 194 open

Rotary embeddings pre-allocate 10x more memory than needed

Agent can lower val_bpb without improving the model — add an optional experiment integrity log?

Proposal: Introduce a Project-Level Long-Term Memory File and a Guidance Agent Mechanism

Is there a plan to reduce hallucinations?

Now with Vision and MoE architecture

Add UCB1 Dimension-Aware Search + Experiment Memory to program.md — no code changes required

Feature: Experiment Novelty Score to Reduce Repeated Search Patterns

RTX 3060 fork — 27 experiments, 25% improvement (1.604 → 1.201 val_bpb)

Codex doesn't seem to work?

Why not fixed FLOPs bugdet?

Request to include Ascend-compatible fork in **Notable forks**

Harden downloaded dataset shard cache in prepare.py

Trust verification for autonomous research agents?

Copilot code review cleanup job fails deleting artifacts with 403

feat: CLI analysis tool for experiment results — structured feedback for the autonomous agent