△ Mechanism confirmed, baseline not beaten
2026
Allocate different entropy budgets to different KV-cache blocks instead of assigning every token and head the same nominal bitwidth. Use the ECASQ Lagrangian so high-variance or attention-sensitive blocks receive more codepoints, while predictable blocks collapse to fewer symbols and become highly compressible. Preserve unbiasedness per scalar or block so reconstructed keys and values have zero mean quantization error conditional on the original tensor.
Useful8/10
Difficulty6/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Replace a square dense projection in a Transformer or MLP with a trainable recursive butterfly matrix. The layer preserves multiscale channel interactions while constraining every complementary row-column block to rank at most k, reducing parameters and enabling recursive structured matrix-vector products. Unlike an arbitrary sparse layer, the construction has an explicit recursive factorization and a quasi-optimal approximation guarantee among matrices with the same butterfly rank.
Useful8/10
Difficulty6/10
Novelty5/10
✗ Failed on benchmark
2026
Replace consecutive or randomly assigned transformed KV coefficients with groups whose variance-volume is approximately equal. Train one equal-size vector-quantizer codebook per group, so a fixed-width cache does not waste its low-rate budget by forcing high-variance and low-variance coordinates into badly mismatched groups. This is a drop-in quantization-layout change that can be applied to keys, values, or both.
Useful8/10
Difficulty5/10
Novelty7/10
✓✓ Beats tuned baseline
2026
Replace dense self-attention by a sparse attention graph whose neighborhoods satisfy the paper's size-dependent expansion condition. This preserves a logarithmically controlled route for every token subset to communicate with the rest of the sequence, reducing quadratic attention cost without allowing disconnected or poorly mixed token groups.
Useful8/10
Difficulty5/10
Novelty5/10
△ Mechanism confirmed, baseline not beaten
2026
Implement causal linear attention in chunks and combine chunk summaries with an associative scan instead of carrying the recurrent state through all chunks sequentially. This preserves the exact causal computation while reducing inter-chunk dependency depth from the number of chunks to its logarithm, enabling substantially more GPU parallelism for long-context training and prefill.
Useful8/10
Difficulty6/10
Novelty6/10
✗ Failed on benchmark
2026
Replace the usual random or elementwise-positive linear-attention feature map with a rank-one positive-semidefinite feature map derived from query and key vectors. For normalized inputs, the resulting kernel is the squared inner product, which is nonnegative and gives a geometrically structured interference pattern that is better suited to associative recall than an arbitrary low-rank feature map.
Useful8/10
Difficulty6/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Replace a dense directed attention matrix by a collection of K learned source-to-hub-to-target interactions. Each hub corresponds to a directed biclique, allowing many source tokens to communicate with many target tokens using O(NK) rather than O(N^2) pair interactions. The construction preserves asymmetric information flow and can be initialized from a graph cover of high-attention edges.
Useful8/10
Difficulty6/10
Novelty6/10
△ Mechanism confirmed, baseline not beaten
2026
Compress each layer's KV tensor with a partial Tucker approximation over token and feature axes, then encode the truncation residual with a rotated uniform quantizer. Select token rank, feature rank, and residual bit-width jointly under a global byte budget, allowing values with flat spectra to receive residual bits while keys may receive more low-rank capacity.
Useful8/10
Difficulty6/10
Novelty6/10
△ Mechanism confirmed, baseline not beaten
2026
Compress a causal sequence by retaining history positions at equal increments of cumulative representation variation instead of at uniform time intervals. Use the resulting N representatives in a decoder that reconstructs piecewise-constant keys and values; the paper's minimax result predicts a worst-case reconstruction error of total variation divided by 2N, independent of where rapid changes occur.
Useful8/10
Difficulty5/10
Novelty6/10
△ Mechanism confirmed, baseline not beaten
2026
Replace token-by-token KV storage after an SSM or recurrent encoder with an online allocate-on-novelty cache. A new slot is created only when the incoming key is sufficiently dissimilar from every stored key; otherwise the incoming value is merged into its nearest slot, so repeated or redundant content does not grow the cache.
Useful8/10
Difficulty4/10
Novelty6/10
✗ Mechanism failed
2026
Use a spherical-design codebook and the paper's polar slack factorization to create a nonnegative geometric interaction bias for attention or expert routing. The resulting kernel is generated by a rank-one term and a rank-at-most-d term, and entries close to zero can define a structured sparse mask instead of relying only on learned top-k selection.
Useful7/10
Difficulty6/10
Novelty7/10
✗ Mechanism failed
2026
Replace the usual JL dimension choice with a task-specific dimension budget based on the paper's asymptotic ranking law. For retrieval, nearest-neighbor search, or compressed attention keys, choose m from a target Kendall correlation and automatically expand the projection when measured ranking quality falls below the target.
Useful7/10
Difficulty3/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Replace dense or single-dilation sparse attention with two sequential sparse attention stages whose offsets form co-prime arithmetic progressions. The first stage mixes tokens separated by multiples of M2, the second by multiples of M1; their composition reaches virtual offsets mM2+nM1, providing many structured long-range interactions from only M1+M2-1 physical offset families. Use causal masking and residual connections so the module can replace a standard transformer attention block without…
Useful7/10
Difficulty5/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Support conjunction queries over multiple roles without explicitly storing a huge tensor of repeated objects. Represent the required higher-order memory through query-dependent contractions, enabling compositional retrieval with memory that scales linearly in the number of objects.
Useful7/10
Difficulty6/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Replace flat all-pairs attention with attention neighborhoods induced by a compatible tree over tokens, patches, nodes, or retrieved items. Retain exact or approximate attention inside nearby tree subtrees and add a path-monotonicity regularizer so semantically distant endpoints are not more similar than intervening tree neighbors.
Useful7/10
Difficulty5/10
Novelty6/10
△ Mechanism confirmed, baseline not beaten
2026
Replace an arbitrary graph-attention mask with a fractional edge mask lying in the intersection of the spanning-tree polytope and twice the matching polytope. The mask represents a distribution over connected spanning trees while imposing expected degree at most two at every vertex, after which sampled trees can be used for sparse message passing.
Useful7/10
Difficulty7/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Replace full PSD self-attention with a pivoted Cholesky/Nyström approximation whose landmarks are sampled from the unexplained diagonal mass. Tokens with large residual self-similarity are more likely to become landmarks, so the rank budget is spent on difficult regions rather than uniformly selected tokens.
Useful7/10
Difficulty5/10
Novelty5/10
△ Mechanism confirmed, baseline not beaten
2026
Represent a batch of token or feature directions as columns of a matrix X, and construct a complementary feature basis Y whose columns are annihilated by X under a diagonal gauge. Use Y as a second algebraically complementary channel for attention or token mixing, either replacing redundant feature projections or regularizing them toward an exact nullspace relation.
Useful7/10
Difficulty6/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Train a fixed-rank neural weight update Y=USV^T with a projector-splitting Runge–Kutta step instead of independently applying Adam or gradient descent to U, S, and V. The update evolves the full low-rank matrix using the neural gradient but performs QR-based factor updates, avoiding S^{-1} and remaining stable when adapter singular values collapse or cross zero. Use a common-base midpoint construction so every internal stage starts from the same U,V basis and remains rank r.
Useful7/10
Difficulty5/10
Novelty6/10
✗ Mechanism failed
2026
Prune redundant attention heads using separate similarity scores for sink behavior and content routing. Two heads are considered safely redundant only when their normalized content compositions are close in Aitchison distance and their sink-mass trajectories are also close, avoiding pruning decisions dominated by a shared sink token.
Useful7/10
Difficulty4/10
Novelty7/10
✗ Failed on benchmark
2026
Insert a linear Johnson–Lindenstrauss bottleneck around a set of jointly processed representations, choosing its width from the sharp finite-set dimension bound rather than from the model's nominal hidden size. The projection should preserve pairwise distances between tokens, patches, or retrieved items, allowing a downstream attention or MLP block to operate at lower width while retaining the geometry relevant to similarity computations.
Useful7/10
Difficulty5/10
Novelty5/10
✗ Failed on benchmark
2026
Replace a wide collection of interchangeable near-zero branches with a module whose output is explicitly a quadratic form in the branch-weight Gram matrix. The module preserves the paper's leading-order behavior while making the relevant collective variable explicit and allowing low-rank parameterizations.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Mechanism failed
2026
Replace dense cross-attention weights with a balanced transport plan whose nonzero query-key edges are maintained by a multiscale active-set procedure. Solve the coarse token-group problem first, lift its support to the fine token grid, add only edges indicated by local cost or marginal residuals, and warm-start the fine problem from the lifted plan. This should provide a principled sparse attention pattern rather than fixing a global top-k pattern before seeing the transport solution.
Useful7/10
Difficulty7/10
Novelty6/10
△ Mechanism confirmed, baseline not beaten
2026
Freeze a randomly initialized single-layer transformer and use a constructed soft prompt to make its attention weights equal Gaussian-kernel weights over support examples. The resulting model performs Nadaraya-Watson regression in one forward pass, so task adaptation stores prompt tokens rather than modifying network weights. Prompt length becomes the number of kernel centers, while hidden dimension and prompt norm determine whether the required logits can be represented accurately.
Useful7/10
Difficulty5/10
Novelty7/10