Search papers, labs, and topics across Lattice.
This paper introduces Effective Feedback Compute (EFC), a metric that quantifies the informativeness, validity, and retention of feedback within agent harnesses, normalizing it by task demand. They demonstrate that EFC significantly outperforms raw compute metrics and a strong SAS baseline in predicting failure rates across various tasks, including synthetic, code execution, and real-world benchmarks. The results highlight that the efficiency of converting raw compute into durable and task-sufficient feedback is the key driver of harness scaling, rather than the sheer amount of computation.
Stop blindly throwing compute at agent harnesses: effective feedback, not raw tokens, dictates scaling laws.
Agent harnesses increasingly determine the performance of language-model systems by deciding how models call tools, receive feedback, verify intermediate states, store memory, and revise solutions. Yet current test-time scaling analyses often parameterize this process by raw expenditure -- tokens, tool calls, operations, wall time, or cost -- which does not distinguish useful feedback from redundant or unstable interaction. We introduce \emph{Effective Feedback Compute} (EFC), a trace-level scaling coordinate that credits feedback only when it is informative, valid, non-redundant, and retained for subsequent decisions, and we normalize it by task demand when comparing tasks with different feedback requirements. Across synthetic controllable tasks, executable code tasks, real benchmark traces, held-out splits, and a prospective validation batch, EFC-based coordinates consistently predict failure rates better than raw-compute baselines and a strong multivariate SAS baseline. In controlled scaling, raw tokens and tool calls explain limited variation ($R^2=0.33$ and $0.42$), SAS reaches $0.88$, while Oracle-EFC and Estimated-EFC reach $0.94$ and Oracle-EFC/$D_{\mathrm{task}}$ reaches $0.99$. Matched-budget interventions show that improving feedback quality raises success from $0.27$ to $0.90$ while raw cost and tool calls are fixed. On mixed real traces, NRS-EFC/$D_{\mathrm{task}}$ reaches $R^2=0.92$ while raw compute has near-zero or negative fit, and it remains the best predictor in a prospective holdout ($R^2=0.85$). These results suggest that harness scaling is governed less by how much computation is spent than by how efficiently raw budget is converted into durable, task-sufficient feedback.
