Shuai Zhang, George Karypis, Xiang Song, Bernie Wang, Jiading Gai
12 of 13 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 11 failures of our instrument; 1 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.
"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2605.15422.json · badge: /api/ran/2605.15422.svg
Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.
| Repository | Role | Ran |
|---|---|---|
| amazon-science/dualkv-flash-attn-for-rl | canonical (classifier label) classifier basis: census |
12 ran · 1 no recorded run |
| Function | Status | Where it lives |
|---|---|---|
| apply_rotary_emb_torch | Ran ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT) |
amazon-science/dualkv-flash-attn-for-rl/flash-attention/flash_attn/layers/rotary.py pointer only (licence: NOASSERTION) · get_code("5ec949977c334f87") How it ranThe body of this sample is withheld: GitHub could not classify the licence of this paper's repository, amazon-science/dualkv-flash-attn-for-rl (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| baichuan_config_to_gpt2_config | Ran ran; our contract draft was wrong, not the code |
amazon-science/dualkv-flash-attn-for-rl/flash-attention/flash_attn/models/baichuan.py pointer only (licence: NOASSERTION) · get_code("ae22707df86d95e3") How it ranThe body of this sample is withheld: GitHub could not classify the licence of this paper's repository, amazon-science/dualkv-flash-attn-for-rl (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| convert_blockmask | Ran ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT) |
amazon-science/dualkv-flash-attn-for-rl/flash-attention/flash_attn/flash_blocksparse_attn_interface.py pointer only (licence: NOASSERTION) · get_code("cc628b160e090ba1") How it ranThe body of this sample is withheld: GitHub could not classify the licence of this paper's repository, amazon-science/dualkv-flash-attn-for-rl (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| flash_attn_qkvpacked_func | Ran ran; our contract draft was wrong, not the code |
amazon-science/dualkv-flash-attn-for-rl/flash-attention/flash_attn/flash_attn_interface.py pointer only (licence: NOASSERTION) · get_code("7541ce9614862680") How it ranThe body of this sample is withheld: GitHub could not classify the licence of this paper's repository, amazon-science/dualkv-flash-attn-for-rl (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| init_to_zero | Ran ran; our contract draft was wrong, not the code |
amazon-science/dualkv-flash-attn-for-rl/flash-attention/flash_attn/flash_attn_triton.py pointer only (licence: NOASSERTION) · get_code("1586b639a5abd0b0") How it ranThe body of this sample is withheld: GitHub could not classify the licence of this paper's repository, amazon-science/dualkv-flash-attn-for-rl (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| maybe_contiguous | Ran ran; our contract draft was wrong, not the code |
amazon-science/dualkv-flash-attn-for-rl/flash-attention/flash_attn/flash_attn_interface.py pointer only (licence: NOASSERTION) · get_code("19ad2ccd82e378c1") How it ranThe body of this sample is withheld: GitHub could not classify the licence of this paper's repository, amazon-science/dualkv-flash-attn-for-rl (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| pad_input | Ran ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT) |
amazon-science/dualkv-flash-attn-for-rl/flash-attention/flash_attn/bert_padding.py pointer only (licence: NOASSERTION) · get_code("a8cf6124094beaab") How it ranThe body of this sample is withheld: GitHub could not classify the licence of this paper's repository, amazon-science/dualkv-flash-attn-for-rl (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| remap_state_dict_hf_baichuan | Ran ran; our contract draft was wrong, not the code |
amazon-science/dualkv-flash-attn-for-rl/flash-attention/flash_attn/models/baichuan.py pointer only (licence: NOASSERTION) · get_code("b83af9adbdeec661") How it ranThe body of this sample is withheld: GitHub could not classify the licence of this paper's repository, amazon-science/dualkv-flash-attn-for-rl (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| rotate_half | Ran ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT) |
amazon-science/dualkv-flash-attn-for-rl/flash-attention/flash_attn/layers/rotary.py pointer only (licence: NOASSERTION) · get_code("163292f43a3c861b") How it ranThe body of this sample is withheld: GitHub could not classify the licence of this paper's repository, amazon-science/dualkv-flash-attn-for-rl (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| round_multiple | Ran ran, honoured the contract we drafted |
amazon-science/dualkv-flash-attn-for-rl/flash-attention/flash_attn/flash_attn_interface.py pointer only (licence: NOASSERTION) · get_code("a0595f1c95cacf06") How it ranThe body of this sample is withheld: GitHub could not classify the licence of this paper's repository, amazon-science/dualkv-flash-attn-for-rl (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| unpad_input | Ran ran; our contract draft was wrong, not the code |
amazon-science/dualkv-flash-attn-for-rl/flash-attention/flash_attn/bert_padding.py pointer only (licence: NOASSERTION) · get_code("d84093c20a0ed8fd") How it ranThe body of this sample is withheld: GitHub could not classify the licence of this paper's repository, amazon-science/dualkv-flash-attn-for-rl (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| unpad_input_for_concatenated_sequences | Ran ran; our contract draft was wrong, not the code |
amazon-science/dualkv-flash-attn-for-rl/flash-attention/flash_attn/bert_padding.py pointer only (licence: NOASSERTION) · get_code("62394c842aafb022") How it ranThe body of this sample is withheld: GitHub could not classify the licence of this paper's repository, amazon-science/dualkv-flash-attn-for-rl (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| setup_fa2 | Not yet run unverified (harvested, no recorded run) |
amazon-science/dualkv-flash-attn-for-rl/flash-attention/benchmarks/benchmark_attn.py pointer only (licence: NOASSERTION) · get_code("a264f868efe9ee7a") How it ranThe body of this sample is withheld: GitHub could not classify the licence of this paper's repository, amazon-science/dualkv-flash-attn-for-rl (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded. |
Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 10 are also present in other papers whose code we hold (10 with at least one paper dated earlier, 0 the same month, 0 later) and 3 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).
Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.
| Function | Also present in | Earliest-dated | Others named |
|---|---|---|---|
| init_to_zero | 12 other papers 12 dated earlier |
2304.11277 (2023-04) PyTorch FSDP: Experiences on Scaling Fully Sharded Data Parallel |
2309.06180 (2023-09) Efficient Memory Management for Large Language Model Serving with PagedAttention 2310.05869 (2023-10) HyperAttention: Long-context Attention in Near-Linear Time and 9 more not named here |
| apply_rotary_emb_torch | 6 other papers 6 dated earlier |
2309.06180 (2023-09) Efficient Memory Management for Large Language Model Serving with PagedAttention |
2407.08608 (2024-07) FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision 2412.06464 (2024-12) Gated Delta Networks: Improving Mamba2 with Delta Rule and 3 more not named here |
| rotate_half | 6 other papers 6 dated earlier |
2309.06180 (2023-09) Efficient Memory Management for Large Language Model Serving with PagedAttention |
2407.08608 (2024-07) FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision 2412.06464 (2024-12) Gated Delta Networks: Improving Mamba2 with Delta Rule and 3 more not named here |
| pad_input | 4 other papers 4 dated earlier |
2309.06180 (2023-09) Efficient Memory Management for Large Language Model Serving with PagedAttention |
2407.08608 (2024-07) FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision 2502.19613 (2025-02) Self-rewarding correction for mathematical reasoning and 1 more not named here |
| unpad_input | 4 other papers 4 dated earlier |
2309.06180 (2023-09) Efficient Memory Management for Large Language Model Serving with PagedAttention |
2407.08608 (2024-07) FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision 2502.19613 (2025-02) Self-rewarding correction for mathematical reasoning and 1 more not named here |
| unpad_input_for_concatenated_sequences | 4 other papers 4 dated earlier |
2309.06180 (2023-09) Efficient Memory Management for Large Language Model Serving with PagedAttention |
2407.08608 (2024-07) FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision 2502.19613 (2025-02) Self-rewarding correction for mathematical reasoning and 1 more not named here |
| convert_blockmask | 3 other papers 3 dated earlier |
2309.06180 (2023-09) Efficient Memory Management for Large Language Model Serving with PagedAttention |
2407.08608 (2024-07) FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision 2604.23798 (2026-04) ELSA: Exact Linear-Scan Attention for Fast and Memory-Light Vision Transformers |
| flash_attn_qkvpacked_func | 3 other papers 3 dated earlier |
2309.06180 (2023-09) Efficient Memory Management for Large Language Model Serving with PagedAttention |
2407.08608 (2024-07) FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision 2604.23798 (2026-04) ELSA: Exact Linear-Scan Attention for Fast and Memory-Light Vision Transformers |
| maybe_contiguous | 3 other papers 3 dated earlier |
2309.06180 (2023-09) Efficient Memory Management for Large Language Model Serving with PagedAttention |
2407.08608 (2024-07) FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision 2604.23798 (2026-04) ELSA: Exact Linear-Scan Attention for Fast and Memory-Light Vision Transformers |
| round_multiple | 3 other papers 3 dated earlier |
2309.06180 (2023-09) Efficient Memory Management for Large Language Model Serving with PagedAttention |
2407.08608 (2024-07) FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision 2604.23798 (2026-04) ELSA: Exact Linear-Scan Attention for Fast and Memory-Light Vision Transformers |
Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
Modern RL post-training methods such as GRPO and DAPO train on N response sequences of R tokens sampled from a shared prompt of P tokens, but standard FlashAttention replicates all P prompt tokens N times across both forward and backward passes -duplicating compute and memory on identical hidden states. In large-rollout, long-context RL training (N ≥16, P ≥8K), this redundancy dominates the policy update cost. We observe that in decoder-only models, causal masking makes prompt representations invariant across sequences at every layer, so all pertoken operations (norms, projections, MLP) and attention can process the prompt once -a property not yet exploited at the kernel level for training. We propose DualKV, the first FlashAttention kernel variant that eliminates shared-prompt replication during RL training, via (1) fused CUDA forward and backward kernels that iterate over two disjoint KV regions -shared context and per-sequence response -in a single kernel launch, and (2) a data-pipeline redesign in veRL that repacks N (P +R) tokens into P +N R tokens per micro-batch, extending the token reduction from attention to the entire model by a factor ρ = N (P +R)/(P +N R). DualKV is mathematically equivalent to standard attention and introduces no approximation. On Qwen3-8B GRPO training with 8×H100 GPUs (N =32, 8Kcontext), DualKV achieves 1.63-2.09× policy-update speedup, enables 2× larger micro-batches, and raises MFU from 36% to 76%. Similar gains hold for DAPO (2.47× speedup, 77% MFU). At 30B MoE scale on 16×H100, DualKV achieves 3.82× policy-update and 3.38× end-to-end step speedup over FlashAttention (which requires 4-way Ulysses sequence parallelism to avoid OOM). DualKV also extends to hybrid sliding/global attention with head dimension 512 (which FA2 does not support) and integrates with Ulysses sequence parallelism, demonstrated on Gemma-4-31B GRPO at 64K context. Modern RL post-training methods such as GRPO [Shao et al., 2024] and DAPO [Yu et al., 2025] compute log-probabilities and gradients for N response sequences that share a prompt of P tokens. With standard FlashAttention-2 (FA2) [Dao, 2024], the training micro-batch contains N sequences of length S i = P + R i , repeating the P -token prompt N times. Each forward and backward pass recomputes K, V activations and their gradients over (N -1) × P redundant prompt tokens per layer. Recent work demonstrates that model accuracy continues to improve with increasingly large rollout factors. Lightman et al. [2024] show that best-of-N performance on the MATH dataset scales log- † Work done while at AWS.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2605.15422")
get_code_for_paper("2605.15422")
have("2605.15422")
The run record, dated, one paper per request, free:
curl https://syntology.ai/api/ran/2605.15422.json
A badge for a README (the split and the date, never a ratio):
[](https://syntology.ai/paper/2605.15422)
Connect an agent — have() is free.