# Response-Sufficient Neural Memory

- ID: 2616
- Canonical URL: https://synthcore.org/idea/2616/response-sufficient-neural-memory
- API JSON: https://synthcore.org/api/idea/2616.json
- API Markdown: https://synthcore.org/api/idea/2616.md
- Verification status: failed_benchmark
- Source: [arXiv:2608.20328](https://arxiv.org/abs/2608.20328)
- Category: memory
- Solves: memory, accuracy, stability
- ML areas: rnn, ssm, world-model
- Math tags: stochastic-processes, statistics, information-theory, dynamical-systems
- Ratings: usefulness 8/10; difficulty 6/10; novelty 8/10

## Idea description

Replace correlation-based memory pruning in an RNN or state-space model by measuring how hidden-state history changes the response to individual past input events. Train a compressed memory coordinate only if it preserves the event-consequence kernel for the target observable, such as future loss, prediction, or control return. A memory representation is accepted when the conditional variance of this kernel within compressed-state groups is small, even if dwell-time or autocorrelation statistics remain large.

## Mathematical statement

For an event record with counting increment \(\mathrm{d}n_a(t)\), where \(a\) indexes an input or internal event, let \(\gamma_a(t)\) be its history-conditioned intensity and define the martingale residual \(\mathrm{d}\varepsilon_a(t)=\mathrm{d}n_a(t)-\gamma_a(t)\mathrm{d}t\). For a future scalar observable \(Y_T\), define the event-consequence kernel \(K_a(t)=\mathbb{E}[Y_T\mid\mathcal{H}_{t-},\mathrm{d}n_a(t)=1]-\mathbb{E}[Y_T\mid\mathcal{H}_{t-},\mathrm{d}n_a(t)=0]\), where \(\mathcal{H}_{t-}\) is the realized history. The trajectory fluctuation expansion is \(Y_T-\mathbb{E}Y_T=\sum_a\int_0^T K_a(t)\,\mathrm{d}\varepsilon_a(t)\). If \(z_t=\phi(\mathcal{H}_t)\) is a compressed neural memory, define the response-heterogeneity gap \(G=\mathbb{E}[\sum_a\int_0^T\gamma_a(t)\operatorname{Var}(K_a(t)\mid z_t,a)\,\mathrm{d}t]\ge 0\). Response sufficiency means that \(G\) is close to zero.

## Key formulas

- $$\mathrm{d}n_a(t)=\gamma_a(t)\,\mathrm{d}t+\mathrm{d}\varepsilon_a(t),\qquad \mathbb{E}[\mathrm{d}\varepsilon_a(t)\mid\mathcal{H}_{t-}]=0,\qquad \mathbb{E}[\mathrm{d}\varepsilon_a(t)\mathrm{d}\varepsilon_b(t)\mid\mathcal{H}_{t-}]=\delta_{ab}\gamma_a(t)\,\mathrm{d}t.$$
- $$K_a(t)=\mathbb{E}[Y_T\mid\mathcal{H}_{t-},\mathrm{d}n_a(t)=1]-\mathbb{E}[Y_T\mid\mathcal{H}_{t-},\mathrm{d}n_a(t)=0].$$
- $$Y_T-\mathbb{E}[Y_T]=\sum_a\int_0^T K_a(t)\,\mathrm{d}\varepsilon_a(t),\qquad G=\mathbb{E}\left[\sum_a\int_0^T\gamma_a(t)\operatorname{Var}(K_a(t)\mid z_t,a)\,\mathrm{d}t\right]\ge 0.$$

## Implementation notes

1. Integration point: apply this to an RNN, GRU, or selective state-space model processing event streams, token sequences, or irregular sensors. Let the hidden state be \(h_t\), and add a bottleneck memory \(z_t=\phi_\psi(h_t)\) with dimension much smaller than \(h_t\). Choose a future observable \(Y_T\), such as validation loss over the next horizon, a future logit, cumulative reward, or a forecast at lag \(\tau\). 2. Pseudocode: sample a trajectory and record event type \(a\), time \(t\), hidden state \(h_t\), and target \(Y_T\). Create a paired counterfactual by masking or delaying the event, or use automatic differentiation through the recurrent update, to estimate \(K_a(t)\approx Y_T(\text{event})-Y_T(\text{no event})\). Predict intensity \(\hat\gamma_a(t)\) from \(h_t\). Train \(\phi_\psi\) and an auxiliary predictor \(q_\omega(z_t,a)\) with weighted loss \(\hat\gamma_a(t)(K_a-q_\omega)^2\), while retaining the task loss. 3. The decomposition and nonnegativity are theoretical; \(K_a\), \(\gamma_a\), and \(G\) are estimated using interventions, Monte Carlo rollouts, or truncated backpropagation. Use cross-fitting for evaluation. 4. First experiment: train a GRU on synthetic delayed-copy event sequences containing one highly autocorrelated irrelevant variable and one weakly correlated but target-relevant variable. Compare full memory, autocorrelation-pruned memory, and response-sufficient memory. The prediction is that long-horizon error tracks \(\hat G\), not autocorrelation; the response method should achieve at least a fivefold lower kernel variance at equal bottleneck size.

## Verification

- Status: failed_benchmark
- Mechanism evidence: yes
- Mechanism confirmed: no
- Practical verdict: harms
- Verdict: Built a response-sufficient memory bottleneck and an analytic Gaussian-Poisson event-stream verification. The martingale residual check had 0.06% relative variance error; the estimated response-heterogeneity gap followed the predicted lambda^2 scaling with 0.19% median relative error and the predicted (1-rho^2) scaling with 0.56% median relative error. At rho=0.35, the response-aware coordinate reduced held-out kernel MSE from 0.885 for the nuisance/correlation proxy to approximately 4e-22, demonstrating the intended mechanism. The future observable MSE itself did not improve because this toy includes irreducible Poisson noise and does not train a recurrent predictor.

### Mechanism check

- Verdict: Built a response-sufficient memory bottleneck and an analytic Gaussian-Poisson event-stream verification. The martingale residual check had 0.06% relative variance error; the estimated response-heterogeneity gap followed the predicted lambda^2 scaling with 0.19% median relative error and the predicted (1-rho^2) scaling with 0.56% median relative error. At rho=0.35, the response-aware coordinate reduced held-out kernel MSE from 0.885 for the nuisance/correlation proxy to approximately 4e-22, demonstrating the intended mechanism. The future observable MSE itself did not improve because this toy includes irreducible Poisson noise and does not train a recurrent predictor.
- Confidence: 9/10
- Limitations: The implementation uses an analytically available event-consequence kernel rather than counterfactual rollouts or automatic differentiation. It is a scalar supervised bottleneck demonstration, not a trained GRU/SSM on delayed-copy sequences; no wall-clock, memory-size, or end-to-end long-horizon task win was established. The binned conditional-gap estimator has finite-bin bias, especially near rho=0.99.

### Practical benchmark

- Paired seeds: 8
- Baseline mean: 0
- Idea mean: 0
- p-value: 0.0081
- Paired wins: 0/8
- Benchmark verdict: idea worse (significant)

## Artifacts

- [bench_report.json](https://synthcore.org/code/982/bench_report.json)
- [experiment.py](https://synthcore.org/code/982/experiment.py)
- [report.md](https://synthcore.org/code/982/report.md)
- [report_bench_2026-09-01T172800.md](https://synthcore.org/code/982/report_bench_2026-09-01T172800.md)
- [response_memory.py](https://synthcore.org/code/982/response_memory.py)
- [results.json](https://synthcore.org/code/982/results.json)
- [stage2_response_memory.py](https://synthcore.org/code/982/stage2_response_memory.py)
- [Download all files as ZIP](https://synthcore.org/download/982)

## Disclaimer

AI-generated research hypothesis, automatically tested. Not peer-reviewed.