Knowledge catalogue
arXiv:2610.02974v1 Announce Type: cross Abstract: Image-based traversability estimation is inherently dependent on the robot platform, deployment domain, and mission preferences, which limits the appl
arXiv:2610.02300v1 Announce Type: new Abstract: Training-free safeguards for text-to-image generation often rely on a reusable safety signal, such as an unsafe direction or global toxic subspace, appl
common : prepare load_from_models_dir() for path conversion (#29674) This is part of the fs::path modernization series. That was also the opportunity to remove fs_list(). Signed-off-by: Adrien Gallouë
spec : fix n-gram drafts rejected at temp > 0 after truncation (#29924) Co-authored-by: Pranesh Gonegandla pgonegandla@nvidia.com Website: https://llama.app Attestations: https://github.com/ggml-org/l
imatrix: calculate activation-based statistics for new format (GGUF) imatrices (#14891) Use activations to calculate the stats Determine calculation mode Compute entropy for activations Compute cosine
cuda : move blocks_per_col to where it is used (#29939) Signed-off-by: Adrien Gallouët angt@huggingface.co Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/5
chat-peg-parser : clear current_tool when pending_tool_call is reset (#29942) A TOOL_ID node that arrives after TOOL_CLOSE wrote through current_tool, which still pointed into the just-destroyed pendi
cuda : move neu_padded to where it is used (#29940) Signed-off-by: Adrien Gallouët angt@huggingface.co Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/52636
ggml-cpu: support BF16/FP16/FP32 K tails in tinyBLAS on x86 (#29806) ggml-cpu: vectorize BF16 K tails in tinyBLAS tests: Skip tinyBLAS when use_ref is enabled so CPU tests compare against the vec_dot
metal : add tensor API flash attention kernel for F16 KV (#29570) metal : add tensor API flash attention kernel for F16 KV cont : add tensor FA kernels for DK=DV=512 and DK=576, DV=512 cont : support
ggml-cpu : fix soft_max_back wrong output when dst aliases src1 (#27096) ggml-cpu : fix soft_max_back wrong output when dst aliases src1 GGML_OP_SOFT_MAX_BACK is listed in ggml_op_can_inplace, so the
ggml-quants : avoid invalid rounding in qkx3 scale search (#29817) ggml-quants : avoid invalid rounding in qkx3 scale search The imatrix scale search can produce an infinite, NaN, or otherwise out-of-
spec : add probabilistic sampling for simple draft and MTP (#27694) Make the drafter probabilistic and the target verify by rejection sampling Drop stale spec_draft_q before drafting Fallback to argma
CUDA: fuse shared experts into MMVQ (#29184) CUDA: fuse shared experts into MMVQ check if buffer is null move stride_col_dst to fusion args Website: https://llama.app Attestations: https://github.com/
model: add support for clef decision model (text-only) (#29831) init support for clef (text only) more static graph clean up nits nits 2 Update gguf-py/gguf/constants.py Co-authored-by: Sigbjørn Skjær
qwen4exp : halve the indexer score memory (#29825) qwen4exp : halve the indexer score memory The indexer scored all heads in one product and rectified a copy of it, so two [n_pool, n_idx_h, n_tokens]
ggml-openvino: update to 2026.4.1, optimize performance, expand ops, improve device listing. (#29852) ggml-openvino : Qwen3.5 MoE perf (#312) Squash of ravi9#312: ggml-openvino: add detailed inference
graph: gather the recurrent states once so the reserve covers every split (#29856) build_rs gathered the extra states (n_rs - n_seqs rows) with their own get_rows. The worst-case reserve has n_rs == n
chat : honor json_schema in Ling 3.0 parser (#29813) chat : honor json_schema in Ling 3.0 parser Ling 3.0 only built a grammar for tool calls and did not handle inputs.json_schema, so response_format
common : add common_is_tty() helper and fix deprecated warnings on Windows (#29860) Signed-off-by: Adrien Gallouët angt@huggingface.co Website: https://llama.app Attestations: https://github.com/ggml-
server : fix laya abort by limiting n_batch to n_ubatch (#29903) server : fix laya abort by limiting n_batch to n_ubatch Fixes #29902 Assisted-by: Claude fix(review) : rm tests, embeddings cond Websit
mtmd : fix deprecated strdup warning on Windows (#29863) Signed-off-by: Adrien Gallouët angt@huggingface.co Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/
Qwen4Exp: add MTP (#29761) Qwen4Exp: add MTP remove has_state member, check via ctx_bufs being non-empty consistent naming + less verbose comments cont : clean-up recurrent memory cont : clean-up comm
CUDA: Handle compute type for NVFP4 on cublass path (#29173) CUDA: Handle compute type for NVFP4 on cublass path Signed-off-by: ynankani ynankani@nvidia.com Use BF16 compute type for quantized models
metal : release temporary private transfer buffers (#29777) metal : release temporary private transfer buffers Assisted-by: OpenAI Codex metal : fix order and formatting Co-authored-by: Niklas Wenzel
cuda : route sm70 to the Turing MMVQ nwarps table (#29753) cuda : route sm70 to the Turing MMVQ nwarps table Volta (sm_70) has no MMVQ parameter table of its own and falls through to GENERIC, which la
hexagon: shared strided DMA copy for CPY and CONCAT, any-dim CONCAT via DMA (#29685) hexagon: shared strided DMA copy for CPY and CONCAT, any-dim CONCAT via DMA hex-cpy: various fixes on top of the co
llama : clamp kpool re-pool bound to existing pools (#29805) tests : simplify function signature llama : clamp kpool re-pool bound to existing pools The n_tokens/kpool + n_seqs_unq bound on n_new_g ov
common,rpc : fix cache dir creation through symlinks on buggy libstdc++ (#29816) See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101510 Close #29759 Signed-off-by: Adrien Gallouët angt@huggingface.co
hexagon: add q2_k and q3_k quant type support (#29717) hexagon: add q2_k and q3_k quant type support hex-qk: consistent allocation of src1_row_size Co-authored-by: Max Krasnyansky maxk@qti.qualcomm.co
hexagon: install rebuilt HTP skels (#29828) hexagon: install rebuilt HTP skels Assisted-by: OpenCode hexagon: fix HTP skel catalog dependencies Assisted-by: OpenCode Website: https://llama.app Attesta
ggml : add alloc_buffer_n to buffer type interface (#23671) ggml : add alloc_buffer_n to buffer type interface Add alloc_buffer_n method to ggml_backend_buffer_type_i interface, with a public API ggml
qwen4exp : optimize mask constructions (#29824) qwen4exp : optimize mask constructions cont : apply the same change for GLM5-next Website: https://llama.app Attestations: https://github.com/ggml-org/l
vulkan: disable large matmul tile on Samsung GPUs with 32KB shared memory (#28531) Assisted-by: Claude Opus Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/
llama, server: add /v1/systemone API (models: laya, julia-1, lev, openjev, kev) (#29818) init conversion convert: ok model loaded add server code improve conversion script support shared prompt prefix
arXiv:2610.01184v1 Announce Type: cross Abstract: Remote multimodal models offer strong numerical reasoning capabilities over charts and speech, but sending private inputs risks exposing sensitive con
Hi folks, I'm a long-time lurker and big fan of this subreddit and a massive self-hosting fan (also outside of AI). I doubt many people will disagree with me here because I see the same arguments bein
llama : preserve original batch order for speculative decoding layer inputs (#29019) llama: preserve original batch order for layer inputs Assisted-by: Codex tests: cover layer-input order across KV l
args: fix cli download mmproj arg (#28977) tests: add tests for cli download arg parsing args: fix cli download mmproj arg Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cp
Hexagon: optimize ALLREDUCE with support for safe scatter mode (#29757) hex-allreduce: add support for safe scatter mode hex-allreduce: pare down excessive comments hex-allreduce: re-write to remove r
cuda: guard the iq4_nl dequantize row kernel against short rows (#29683) dequantize_block_iq4_nl writes QK_K values per block, but a row can be shorter than that (an IQ4_NL row is only guaranteed to b
ggml-opencl : replace alloca() with std::vector (#29765) Signed-off-by: Adrien Gallouët angt@huggingface.co Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/
model : re-enable -sm tensor for qwen4exp (#28569) #27941 disabled -sm tensor for qwen4exp because test-llama-archs asserted on the Meta device once the fixture carried a PLE layer: GGML_ASSERT(ggml_b
llama-bench : fix verbosity filter to show GGML_LOG_ERROR (#28229) bench : fix verbosity filter to show GGML_LOG_ERROR (#28107) bench: remove dead variables Website: https://llama.app Attestations: ht
vocab : honor BOS/EOS settings for PLaMo-2 and PLaMo-3 (#29734) vocab : honor BOS/EOS settings for PLaMo-2 and PLaMo-3 The original tokenizer configs for PLaMo-2 and PLaMo-3 have add_bos_token: true a
common : add LLM-jp-4.1 Harmony dialect handler (#29681) LLM-jp-4.1 uses the GPT-OSS format, but its tokenizer decodes a space after every special token and parallel tool calls are separated by <|end|
BLAS : Document AOCL-BLAS build and label the device AOCL-BLAS (#29640) BLAS : Document AOCL-BLAS build and label the device AOCL-BLAS AOCL-Blas : Add an AOCL-BLAS Quick Start and drop the fixed versi
hex-workqueue: fix race condition in seqn getting out of sync with idx_read/write (#29785) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/51906249 macOS/iO
HIP: avoid treating CDNA as dgx spark for gqa_ratio 20 in fattn_mma dqk 576 (#29572) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/51919562 macOS/iOS: mac
llama-mmap : avoid a second full-size copy of each tensor with direct-io (#29749) Assisted-by: Claude Co-authored-by: Pranesh Gonegandla pgonegandla@nvidia.com Website: https://llama.app Attestations:
arXiv:2609.38294v1 Announce Type: new Abstract: We study the generation of agentic workflows that jointly optimize multiple objectives, such as accuracy, cost, latency, robustness, and consistency. Ex
vulkan: Tune GDN kernel, fix Intel performance (#29476) vulkan: tune GDN shader tune for Intel Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/51306598 macO
ci: add zdnn backend build but not test (#29541) ci: add zdnn backend build but not test Signed-off-by: Aaron Teo aaron.teo1@ibm.com ci: attempt to run a ubuntu 26.04 container Signed-off-by: Aaron Te
ggml-cuda: HIP: optimize packed byte subtraction (__vsubss4 -> __vsub4) (#29478) ggml-cuda: HIP: optimize non-saturating packed byte subtraction (__vsubss4) CI: ignore 1 spilled vgpr in fattn_vec Co-a
hexagon: optimize concat op (#29673) hex-concat: reduce pkts in gather/transpose hot loop gather directly into dst buffer, use special instruction for gather sync hex-concat: use fastdiv replace calls
model : support classifier_pooling for rerankers (#29627) model : support classifier_pooling for ModernBERT rerankers Assisted-by: Claude Opus 5.5 model : read classifier pooling type in load_hparams
ggml : check row bounds in get_rows_back (#29575) Signed-off-by: Adrien Gallouët angt@huggingface.co Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/5139241
gguf : reject tensor size that wraps after padding (#26979) GGML_PAD(nbytes, alignment) wraps to 0 when nbytes is within (alignment - 1) of SIZE_MAX, which silently bypassed the size overflow guard in
Hexagon f16 activation ops (#29209) hexagon: add F16 support for activation ops (SILU/GELU/GELU_QUICK/GEGLU/SWIGLU) Widens ggml_hexagon_supported_activations() to accept F16 (src0/dst/src1 must agree
ggml-zdnn: impl buffer reset, fix memory leaks (#29637) cont: fix code style Signed-off-by: Aaron Teo aaron.teo1@ibm.com Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/